speech-seq2seq/ami

1AMI_EN2001a_H00_MEE068_0000557_0000594 Okay.2AMI_EN2001a_H00_MEE068_0001109_0001553 Does anyone want to see uh Steve's feedback from the specification?3AMI_EN2001a_H00_MEE068_0001895_0001926 Right.4AMI_EN2001a_H00_MEE068_0002141_0002577 Not really, um just what he's talking about, like duplication of effort and5AMI_EN2001a_H00_MEE068_0002809_0003300 Like duplication of effort and stuff, and um yeah,6AMI_EN2001a_H00_MEE068_0003300_0003961 he was saying that we should maybe uh think about having a prototype for week six, which is next week.7AMI_EN2001a_H00_MEE068_0003961_0004004 Yeah.8AMI_EN2001a_H00_MEE068_0004224_0004746 So we should probably prioritize our packages.9AMI_EN2001a_H00_MEE068_0005664_0005689 Mm.10AMI_EN2001a_H00_MEE068_0008002_0008038 Yeah.11AMI_EN2001a_H00_MEE068_0012194_0012214 Yeah.12AMI_EN2001a_H00_MEE068_0015547_0015635 Hmm.13AMI_EN2001a_H00_MEE068_0015635_0016140 Has has anyone actually looked at the Java code for the, huh?14AMI_EN2001a_H00_MEE068_0016140_0016169 Hmm.15AMI_EN2001a_H00_MEE068_0018680_0018752 Yeah, I think so.16AMI_EN2001a_H00_MEE068_0025227_0025638 Yeah, I I don't know about the search functionality, that might be online.17AMI_EN2001a_H00_MEE068_0025638_0025750 Depends how it's gonna work.18AMI_EN2001a_H00_MEE068_0026583_0026620 Yeah.19AMI_EN2001a_H00_MEE068_0046098_0046122 Mm-hmm.20AMI_EN2001a_H00_MEE068_0047962_0048040 Yeah, that makes sense.21AMI_EN2001a_H00_MEE068_0057829_0057858 Hmm.22AMI_EN2001a_H00_MEE068_0065547_0065568 Hmm.23AMI_EN2001a_H00_MEE068_0097060_0097237 Yeah, you just concatenate them together.24AMI_EN2001a_H00_MEE068_0110595_0110627 Hmm.25AMI_EN2001a_H00_MEE068_0113146_0113177 Yeah.26AMI_EN2001a_H00_MEE068_0113446_0113626 It just means it loads on demand.27AMI_EN2001a_H00_MEE068_0113626_0114108 It only loads when it needs a particular type of file.28AMI_EN2001a_H00_MEE068_0114108_0114222 Like when it's being accessed.29AMI_EN2001a_H00_MEE068_0116076_0116436 Yeah, I think that's the idea, it just loads the particular ones it needs.30AMI_EN2001a_H00_MEE068_0116750_0117023 But if you were doing a search over the whole corpus you'd have to load them all.31AMI_EN2001a_H00_MEE068_0117919_0117947 Hmm.32AMI_EN2001a_H00_MEE068_0132274_0132288 Mm.33AMI_EN2001a_H00_MEE068_0149028_0149049 Hmm.34AMI_EN2001a_H00_MEE068_0149168_0149521 Yeah, we do not want it in to develop a little tree display as well for multiple results.35AMI_EN2001a_H00_MEE068_0149704_0149864 Yeah, but that'd be quite easy to do.36AMI_EN2001a_H00_MEE068_0150352_0150490 You just need to find the time stamp.37AMI_EN2001a_H00_MEE068_0151024_0151064 Yeah.38AMI_EN2001a_H00_MEE068_0154912_0154940 Yeah.39AMI_EN2001a_H00_MEE068_0163886_0164296 Yeah, I think I think those segments for each utterance are split up.40AMI_EN2001a_H00_MEE068_0164420_0164462 Think so.41AMI_EN2001a_H00_MEE068_0166346_0166540 Yeah, I'm pretty sure it's already there.42AMI_EN2001a_H00_MEE068_0166540_0166638 Pretty sure that's already there.43AMI_EN2001a_H00_MEE068_0166638_0166893 The the utterances are numbered.44AMI_EN2001a_H00_MEE068_0170479_0170505 Hmm.45AMI_EN2001a_H00_MEE068_0171159_0171218 Yeah, I think so.46AMI_EN2001a_H00_MEE068_0174565_0174759 Ye that's the impression I get, yeah.47AMI_EN2001a_H00_MEE068_0175223_0175242 Oh.48AMI_EN2001a_H00_MEE068_0178364_0178386 Hmm.49AMI_EN2001a_H00_MEE068_0178610_0178621 Ye50AMI_EN2001a_H00_MEE068_0178721_0178743 Mm.51AMI_EN2001a_H00_MEE068_0179101_0179160 Yeah, uh52AMI_EN2001a_H00_MEE068_0181418_0181435 Right.53AMI_EN2001a_H00_MEE068_0184072_0184112 Okay.54AMI_EN2001a_H00_MEE068_0185038_0185111 Topics, yeah.55AMI_EN2001a_H00_MEE068_0185896_0185986 Yeah, I think that's the right one.56AMI_EN2001a_H00_MEE068_0187080_0187103 Hmm.57AMI_EN2001a_H00_MEE068_0188804_0188825 Hmm.58AMI_EN2001a_H00_MEE068_0189236_0189262 Mm-hmm.59AMI_EN2001a_H00_MEE068_0202647_0202678 Mm.60AMI_EN2001a_H00_MEE068_0205991_0206009 Hmm.61AMI_EN2001a_H00_MEE068_0210165_0210335 Yeah, that'd be much more efficient to do that.62AMI_EN2001a_H00_MEE068_0210335_0210358 Yeah.63AMI_EN2001a_H00_MEE068_0216956_0216988 Hmm.64AMI_EN2001a_H00_MEE068_0218806_0218816 Hmm.65AMI_EN2001a_H00_MEE068_0225260_0225392 Yeah, you're able to do that in Java, yeah?66AMI_EN2001a_H00_MEE068_0225573_0225596 Yeah.67AMI_EN2001a_H00_MEE068_0227806_0227817 Huh.68AMI_EN2001a_H00_MEE068_0231117_0231148 Hmm.69AMI_EN2001a_H00_MEE068_0232829_0233392 Yeah, I've had a b I've had a look at the the topic segments, how it's stored.70AMI_EN2001a_H00_MEE068_0233392_0233921 And then yeah, th those are few per meeting, and it um well71AMI_EN2001a_H00_MEE068_0233921_0234425 it gives a time stamp and inside each one there's uh the actual like utterance segments.72AMI_EN2001a_H00_MEE068_0234535_0234692 And the list of them that occurred.73AMI_EN2001a_H00_MEE068_0234692_0234773 And they're all numbered.74AMI_EN2001a_H00_MEE068_0234921_0235076 Um so that's where that's stored.75AMI_EN2001a_H00_MEE068_0235832_0235852 Yeah76AMI_EN2001a_H00_MEE068_0235852_0236921 so I guess um if I'm gonna be segmenting it with a L. C. seg then that's like same format I'd want to um put it back out in so it'd be equivalent.77AMI_EN2001a_H00_MEE068_0238071_0238238 Well, like the integration.78AMI_EN2001a_H00_MEE068_0238238_0238335 What do you mean, integration?79AMI_EN2001a_H00_MEE068_0238776_0238823 Hmm.80AMI_EN2001a_H00_MEE068_0239148_0239175 I don't know.81AMI_EN2001a_H00_MEE068_0239175_0239371 I don't think anyone's been allocated to do that yet.82AMI_EN2001a_H00_MEE068_0240098_0240136 Yeah, yeah.83AMI_EN2001a_H00_MEE068_0240325_0240380 Yeah, definitely.84AMI_EN2001a_H00_MEE068_0242021_0242111 Hmm, yeah.85AMI_EN2001a_H00_MEE068_0243162_0243332 Yeah, it c could be difficult, yeah.86AMI_EN2001a_H00_MEE068_0244760_0244794 Yeah.87AMI_EN2001a_H00_MEE068_0244919_0245371 Well I guess the important thing is to get the crucial m modules built.88AMI_EN2001a_H00_MEE068_0245505_0245604 Ye yeah.89AMI_EN2001a_H00_MEE068_0245760_0245844 Yeah, and then90AMI_EN2001a_H00_MEE068_0246067_0246584 Yeah, and then we'll maybe have to prioritize somebody into just integrating it.91AMI_EN2001a_H00_MEE068_0261716_0261742 Mm-hmm.92AMI_EN2001a_H00_MEE068_0262760_0262820 Yeah, I think so.93AMI_EN2001a_H00_MEE068_0264132_0264161 Uh yeah.94AMI_EN2001a_H00_MEE068_0264927_0264997 Hmm.95AMI_EN2001a_H00_MEE068_0264997_0265047 Yeah, yeah.96AMI_EN2001a_H00_MEE068_0265047_0265492 Jasmine, I thought you just said that you'd uh looked at extracting the text.97AMI_EN2001a_H00_MEE068_0266349_0266428 Yeah.98AMI_EN2001a_H00_MEE068_0266428_0266669 So you you said you did it in Python, yeah?99AMI_EN2001a_H00_MEE068_0266879_0267230 Yeah, did you use uh b the X. L. uh X. M. L. parser in Python?100AMI_EN2001a_H00_MEE068_0267230_0267279 Right.101AMI_EN2001a_H00_MEE068_0267279_0267436 Yeah, sounds pretty good.102AMI_EN2001a_H00_MEE068_0267436_0267635 So um 'cause, yeah103AMI_EN2001a_H00_MEE068_0267635_0268354 I was having a look in it a look at it as well and I noticed the um the speakers are all in that separate file?104AMI_EN2001a_H00_MEE068_0268354_0268647 So did did you have to combine them all and and then re-order them?105AMI_EN2001a_H00_MEE068_0268973_0269002 Yeah.106AMI_EN2001a_H00_MEE068_0269131_0269181 Ye yeah, c107AMI_EN2001a_H00_MEE068_0269303_0269414 Right.108AMI_EN2001a_H00_MEE068_0269414_0269749 Yeah, so that's approach um well, I was going to do.109AMI_EN2001a_H00_MEE068_0269749_0269966 So yeah, we may as well collaborate.110AMI_EN2001a_H00_MEE068_0270861_0270928 In the word files?111AMI_EN2001a_H00_MEE068_0272879_0273041 I'm not sure I what you mean.112AMI_EN2001a_H00_MEE068_0276568_0276603 Oh right.113AMI_EN2001a_H00_MEE068_0277636_0277658 Hmm.114AMI_EN2001a_H00_MEE068_0283826_0283842 Hmm.115AMI_EN2001a_H00_MEE068_0285911_0286195 Mm I thought they were local to th a particular meeting.116AMI_EN2001a_H00_MEE068_0344681_0344698 Hmm.117AMI_EN2001a_H00_MEE068_0409941_0410128 Mm is there anything else we should discuss?118AMI_EN2001a_H00_MEE068_0412745_0413295 Yeah, should we not have like a group directory or something where we can put all our code in and that kinda thing?119AMI_EN2001a_H00_MEE068_0413663_0413691 Hmm.120AMI_EN2001a_H00_MEE068_0414915_0415078 I've gotten mm hardly any121AMI_EN2001a_H00_MEE068_0416231_0416262 Hmm.122AMI_EN2001a_H00_MEE068_0417688_0417972 Yeah, we can ask Steve if um we can get space.123AMI_EN2001a_H00_MEE068_0418566_0418700 Yeah, uh we could do that.124AMI_EN2001a_H00_MEE068_0419381_0419568 Yeah, I'm sure he had to deal with that last year.125AMI_EN2001a_H00_MEE068_0425489_0425532 Yeah.126AMI_EN2001a_H00_MEE068_0438747_0438768 Hmm.127AMI_EN2001a_H00_MEE068_0443772_0443916 That sounds good.128AMI_EN2001a_H00_MEE068_0447071_0447099 Hmm.129AMI_EN2001a_H00_MEE068_0452622_0452642 Hmm.130AMI_EN2001a_H00_MEE068_0453472_0453671 Yeah, that's what I'm gonna need.131AMI_EN2001a_H00_MEE068_0454312_0454343 Yes.132AMI_EN2001a_H00_MEE068_0454460_0454603 Yeah, it's just mo changing it a bit.133AMI_EN2001a_H00_MEE068_0455575_0455593 Yeah.134AMI_EN2001a_H00_MEE068_0458710_0458974 No, but uh that's what M. L. C. seg does.135AMI_EN2001a_H00_MEE068_0458974_0459202 It it marks the end of each segment.136AMI_EN2001a_H00_MEE068_0459457_0459476 Yeah.137AMI_EN2001a_H00_MEE068_0460413_0460434 Yeah.138AMI_EN2001a_H00_MEE068_0462659_0462679 Oh.139AMI_EN2001a_H00_MEE068_0463654_0463677 Yeah.140AMI_EN2001a_H00_MEE068_0463820_0464033 Yeah, for me it's better if they're by meeting.141AMI_EN2001a_H00_MEE068_0471576_0471910 Then that'll be really easy to do once they've got the raw text.142AMI_EN2001a_H00_MEE068_0471910_0472132 It's just a case of running the script.143AMI_EN2001a_H00_MEE068_0472485_0472645 Yeah, I mean hopefully this week.144AMI_EN2001a_H00_MEE068_0474692_0474708 Alright.145AMI_EN2001a_H00_MEE068_0476771_0476828 And we could146AMI_EN2001a_H00_MEE068_0503962_0503992 Don't know.147AMI_EN2001a_H00_MEE068_0503992_0504131 Suppose we're just getting on with all our148AMI_EN2001a_H00_MEE068_0504245_0504312 components.149AMI_EN2001a_H00_MEE068_0504312_0504343 So150AMI_EN2001a_H00_MEE068_0504725_0504883 I know.151AMI_EN2001a_H00_MEE068_0504883_0504899 Wa152AMI_EN2001a_H00_MEE068_0505943_0506009 Yeah.153AMI_EN2001a_H00_MEE068_0506009_0506296 Yeah, he suggested that we could have an uh initial prototype.154AMI_EN2001a_H00_MEE068_0506670_0507230 I know, I'd b I'd be surprised if we can get anything working by next week.155AMI_EN2001a_H00_MEE068_0508107_0508124 Alright.156AMI_EN2001a_H01_FEO066_0007881_0007913 Yeah.157AMI_EN2001a_H01_FEO066_0008438_0009432 Yeah, I mean if we just want to have um some data for the user face, could even be random data.158AMI_EN2001a_H01_FEO066_0009432_0009471 Uh mm mm159AMI_EN2001a_H01_FEO066_0010400_0010458 Yeah, I'm160AMI_EN2001a_H01_FEO066_0025303_0025350 Hmm.161AMI_EN2001a_H01_FEO066_0040639_0040685 Yes.162AMI_EN2001a_H01_FEO066_0046127_0046169 Hmm yes.163AMI_EN2001a_H01_FEO066_0047482_0047500 Hmm.164AMI_EN2001a_H01_FEO066_0049430_0049647 I'm not so sure.165AMI_EN2001a_H01_FEO066_0051155_0052234 I I thought we would just have like um one big summary um with all the uh different importance levels um displayed on it.166AMI_EN2001a_H01_FEO066_0052234_0052845 And depending on what our um zoom level is, we just display a part of it.167AMI_EN2001a_H01_FEO066_0053182_0053594 And we would have one very big thing off-line.168AMI_EN2001a_H01_FEO066_0053594_0053859 And from that we would just select what we are displaying.169AMI_EN2001a_H01_FEO066_0054204_0054285 Yes.170AMI_EN2001a_H01_FEO066_0054285_0055276 So for example you would um give a high value to those um sequences you want to display in the meeting series summary.171AMI_EN2001a_H01_FEO066_0055276_0055408 And you just cut off.172AMI_EN2001a_H01_FEO066_0056607_0056837 That was what I sh I thought, yeah.173AMI_EN2001a_H01_FEO066_0056837_0056903 I thought.174AMI_EN2001a_H01_FEO066_0063964_0064517 But I think the m difference might be that we want just want to have um the words.175AMI_EN2001a_H01_FEO066_0064517_0065545 And that's not so much what he meant with not possibly loading everything was that you m um load all the uh annotation stuff,176AMI_EN2001a_H01_FEO066_0065545_0065707 all the sound files, all177AMI_EN2001a_H01_FEO066_0069234_0069385 In178AMI_EN2001a_H01_FEO066_0071564_0073059 Um I r I I'm getting quite lost um at the moment because um w what's um our difference between the um se um uh the importance measure and the skimming?179AMI_EN2001a_H01_FEO066_0073059_0073342 I mean, do we do both or is it the same thing?180AMI_EN2001a_H01_FEO066_0073724_0073763 Okay.181AMI_EN2001a_H01_FEO066_0073900_0074527 So but when when we talk about summaries you talk about this uh abo about skimming and not about182AMI_EN2001a_H01_FEO066_0076251_0076279 Yeah.183AMI_EN2001a_H01_FEO066_0076680_0076899 Yeah right, isn't that the skimming?184AMI_EN2001a_H01_FEO066_0076899_0077006 Isn't that the skimming?185AMI_EN2001a_H01_FEO066_0077427_0077588 Yeah, but it use the same data.186AMI_EN2001a_H01_FEO066_0084215_0084269 Yeah.187AMI_EN2001a_H01_FEO066_0084433_0084485 A188AMI_EN2001a_H01_FEO066_0085529_0086549 And, yeah, I think we also thought about combining that measure with um the measures I get from um s uh hot-spots and so on.189AMI_EN2001a_H01_FEO066_0086549_0087009 So that would also be on utterance level, I think.190AMI_EN2001a_H01_FEO066_0087009_0087067 I think.191AMI_EN2001a_H01_FEO066_0107332_0107405 Yes, sure.192AMI_EN2001a_H01_FEO066_0113111_0113148 Yes.193AMI_EN2001a_H01_FEO066_0117540_0117620 Yes, right.194AMI_EN2001a_H01_FEO066_0118082_0118130 Oops, it does.195AMI_EN2001a_H01_FEO066_0118130_0118441 So I define baseline and what it loads?196AMI_EN2001a_H01_FEO066_0118441_0119776 For example it loads all the utterances and so on, but it doesn't load um the discourse acts and for example not the and what's what else there?197AMI_EN2001a_H01_FEO066_0119776_0119995 Not the summaries.198AMI_EN2001a_H01_FEO066_0119995_0120172 It only loads those on demand.199AMI_EN2001a_H01_FEO066_0125501_0126258 Y you mean that you um basically split up th the big thing into um different summaries.200AMI_EN2001a_H01_FEO066_0126258_0127308 For example that you have a very um top-level um summary and a separate file for for each level.201AMI_EN2001a_H01_FEO066_0130256_0130283 Mm-hmm.202AMI_EN2001a_H01_FEO066_0160427_0160472 Yes.203AMI_EN2001a_H01_FEO066_0162752_0162930 N Uh no no, it's f for204AMI_EN2001a_H01_FEO066_0163256_0163329 No, you're right.205AMI_EN2001a_H01_FEO066_0163329_0163411 Yeah.206AMI_EN2001a_H01_FEO066_0163411_0163490 It's for207AMI_EN2001a_H01_FEO066_0167456_0167506 Um208AMI_EN2001a_H01_FEO066_0169380_0170275 No, I I think we would just take the segments that are already that were Yeah, there's um this segments file.209AMI_EN2001a_H01_FEO066_0170275_0170442 Um you know, the X. M. L. segments.210AMI_EN2001a_H01_FEO066_0170749_0170763 Oh.211AMI_EN2001a_H01_FEO066_0171124_0171213 That I don't know.212AMI_EN2001a_H01_FEO066_0174820_0174923 Yeah, that's um213AMI_EN2001a_H01_FEO066_0175133_0175168 Mm-hmm.214AMI_EN2001a_H01_FEO066_0176870_0177175 There there are time stamps um for, well,215AMI_EN2001a_H01_FEO066_0177175_0178128 segments um and for th um segments is for example when when you look at the data, what is displayed in one line.216AMI_EN2001a_H01_FEO066_0178380_0178695 What when when you look at it in the hmm?217AMI_EN2001a_H01_FEO066_0178961_0179096 I think so.218AMI_EN2001a_H01_FEO066_0179096_0179129 Isn't219AMI_EN2001a_H01_FEO066_0179294_0180027 Um for ex um I I compared it with what I did for the pause um duration extraction.220AMI_EN2001a_H01_FEO066_0180027_0180652 Um and basically it's uh words that are uttered in a sequence without pauses.221AMI_EN2001a_H01_FEO066_0180652_0181632 But sometimes um however there are um short pauses in it and they're indicated by square brackets pause or something in the data.222AMI_EN2001a_H01_FEO066_0182055_0182667 Um someti uh but uh the annotators decided what was one segment and what wasn't.223AMI_EN2001a_H01_FEO066_0183607_0183715 I think so.224AMI_EN2001a_H01_FEO066_0187094_0188191 Yeah, but um I think for some annotations um an uttera ca utterance can have several um types.225AMI_EN2001a_H01_FEO066_0188191_0188384 For example for the dialogue acts and so on.226AMI_EN2001a_H01_FEO066_0188655_0188719 Okay.227AMI_EN2001a_H01_FEO066_0188719_0188844 Yeah, that should be for228AMI_EN2001a_H01_FEO066_0189116_0189239 Yeah.229AMI_EN2001a_H01_FEO066_0189239_0189299 Should be, yeah.230AMI_EN2001a_H01_FEO066_0271077_0271213 Yes, but that's231AMI_EN2001a_H01_FEO066_0271434_0271766 Yeah, everything that's a word has a sti time stamp.232AMI_EN2001a_H01_FEO066_0275950_0276142 That's at the end.233AMI_EN2001a_H01_FEO066_0276142_0276353 That's at the end, I think, her time.234AMI_EN2001a_H01_FEO066_0276696_0276792 Yeah, maybe.235AMI_EN2001a_H01_FEO066_0276924_0277047 Didn't have a look at our meetings.236AMI_EN2001a_H01_FEO066_0277656_0278507 Uh I I think it wouldn't as it occurs I mean it would be it occurs in every meeting.237AMI_EN2001a_H01_FEO066_0278507_0278595 So238AMI_EN2001a_H01_FEO066_0278935_0279426 And I think it even has uh its own annotation, like digits or something.239AMI_EN2001a_H01_FEO066_0279426_0279756 So that should be really easy to cut out.240AMI_EN2001a_H01_FEO066_0280228_0280387 Yeah.241AMI_EN2001a_H01_FEO066_0280387_0280636 I'm sure.242AMI_EN2001a_H01_FEO066_0281097_0281296 Ah it's just to test the system, I think.243AMI_EN2001a_H01_FEO066_0281419_0281431 So244AMI_EN2001a_H01_FEO066_0282105_0282373 Mm they have to read numbers from245AMI_EN2001a_H01_FEO066_0283443_0283566 Uh I didn't have a look at that.246AMI_EN2001a_H01_FEO066_0283566_0283728 So247AMI_EN2001a_H01_FEO066_0285992_0286018 They248AMI_EN2001a_H01_FEO066_0315434_0315459 Mm-hmm.249AMI_EN2001a_H01_FEO066_0333444_0333587 Uh th yeah.250AMI_EN2001a_H01_FEO066_0333826_0333857 'Kay.251AMI_EN2001a_H01_FEO066_0333857_0334962 Um I just um wondered, so who's uh then doing um the frequencies on on the words252AMI_EN2001a_H01_FEO066_0334962_0335843 because I'm I think I will also um I could also make use of it um for the agreement and disagreement thing.253AMI_EN2001a_H01_FEO066_0335843_0336878 Because I um I in my outline I talked about um using the um discourse acts first254AMI_EN2001a_H01_FEO066_0336878_0337667 and um then in the chunks of text I found looking for word patterns and so on.255AMI_EN2001a_H01_FEO066_0337667_0338486 So um I would for example need the um most freq um frequent words.256AMI_EN2001a_H01_FEO066_0338486_0338990 So if you cut off all that, I'd won't be use or257AMI_EN2001a_H01_FEO066_0339910_0340174 Yeah, I I but I need it for my chunks then.258AMI_EN2001a_H01_FEO066_0340174_0340215 I would259AMI_EN2001a_H01_FEO066_0340351_0340384 You know?260AMI_EN2001a_H01_FEO066_0340660_0341896 Yeah, but I'd uh I would like to look at the frequency of words in my um in the regions of text I found out to be interesting.261AMI_EN2001a_H01_FEO066_0341896_0342034 So I wouldn't need it.262AMI_EN2001a_H01_FEO066_0342034_0342401 It it would have to be re-calculated only for my segments.263AMI_EN2001a_H01_FEO066_0342885_0342905 Huh?264AMI_EN2001a_H01_FEO066_0343135_0343375 Uh uh mm.265AMI_EN2001a_H01_FEO066_0343375_0344331 I think it would be, you know, l as as big at as the hot-spot annotation things.266AMI_EN2001a_H01_FEO066_0344331_0344581 That's quite small, yeah, that's some utterances.267AMI_EN2001a_H01_FEO066_0345987_0346037 Yes.268AMI_EN2001a_H01_FEO066_0346037_0346139 Yeah, yeah.269AMI_EN2001a_H01_FEO066_0346139_0346904 So I would probably just concatenate all my um text chunks and then let's say m I will run over it.270AMI_EN2001a_H01_FEO066_0347513_0347554 Yes.271AMI_EN2001a_H01_FEO066_0382337_0382556 Yes, definitely.272AMI_EN2001a_H01_FEO066_0424617_0424759 Yeah, right.273AMI_EN2001a_H01_FEO066_0424934_0425033 Ye274AMI_EN2001a_H01_FEO066_0425228_0425232 M275AMI_EN2001a_H01_FEO066_0425632_0426431 Um Jasmine, uh um what is um the text you're extracting uh looking like then at the end?276AMI_EN2001a_H01_FEO066_0426832_0429243 Because um I I think it's actually very similar to what I did for my um speaker um uh extraction and I think I would ch perhaps have to change two lines of codes to get you um for each meeting a file that says fr from um this millisecond to this millisecond there was this sequence of words.277AMI_EN2001a_H01_FEO066_0429413_0429462 And so on.278AMI_EN2001a_H01_FEO066_0429614_0430014 So that's just changing two lines of code.279AMI_EN2001a_H01_FEO066_0430014_0430122 And it would give you that.280AMI_EN2001a_H01_FEO066_0430122_0430141 So281AMI_EN2001a_H01_FEO066_0430781_0430854 Um yeah.282AMI_EN2001a_H01_FEO066_0430854_0431275 So far I extracted um the dura durations.283AMI_EN2001a_H01_FEO066_0431275_0431468 But it's from the words file.284AMI_EN2001a_H01_FEO066_0431468_0432679 So I could just um contatenate concatenate um the words instead of the durations, and it should I mean285AMI_EN2001a_H01_FEO066_0432933_0433032 Should be very286AMI_EN2001a_H01_FEO066_0433151_0433214 straight-forward.287AMI_EN2001a_H01_FEO066_0433402_0433692 I can try to do it and send it to you.288AMI_EN2001a_H01_FEO066_0433692_0434102 Pe and you have a look at it, will it make sense for what you want.289AMI_EN2001a_H01_FEO066_0435456_0435860 Yeah, uh p I mean it I just let it run over all the files.290AMI_EN2001a_H01_FEO066_0435860_0435884 So291AMI_EN2001a_H01_FEO066_0436106_0436222 Yes.292AMI_EN2001a_H01_FEO066_0436222_0436297 I just ordered.293AMI_EN2001a_H01_FEO066_0436592_0437029 Uh I ordered according to the um starting times of the utterances.294AMI_EN2001a_H01_FEO066_0438218_0438337 What do you mean by diffe295AMI_EN2001a_H01_FEO066_0439271_0439818 Yeah, I mean t I I have one what I give you would be one file for each meeting.296AMI_EN2001a_H01_FEO066_0440157_0440355 Yeah, not for each meeting series.297AMI_EN2001a_H01_FEO066_0440355_0440439 I didn't do that yet.298AMI_EN2001a_H01_FEO066_0441100_0441243 Yeah, one group, yeah.299AMI_EN2001a_H01_FEO066_0443405_0443947 Yeah, I mean there's one series that has just one meeting.300AMI_EN2001a_H01_FEO066_0444488_0444571 Yes.301AMI_EN2001a_H01_FEO066_0444571_0445131 Um the you you the data is of the form you have um three identification letter.302AMI_EN2001a_H01_FEO066_0445131_0445609 So B. E. D. or B. B. D. or something, and that's always the same group.303AMI_EN2001a_H01_FEO066_0445609_0446096 And then after that there's um a number like O. O. one, O. O. two.304AMI_EN2001a_H01_FEO066_0446096_0446185 So, it's a305AMI_EN2001a_H01_FEO066_0447022_0447528 Yeah, but that's that's really quite easy to see because they're named.306AMI_EN2001a_H01_FEO066_0449210_0449300 Yes.307AMI_EN2001a_H01_FEO066_0449300_0450158 But I I mean as um the start uh start times um start for each meeting at zero308AMI_EN2001a_H01_FEO066_0450158_0451220 you could just probably just um add the um final second time to the next meeting and so on and just put it all together.309AMI_EN2001a_H01_FEO066_0451220_0452218 But then we would have to change um the information about who on which channel it was set, um to by which person it was set.310AMI_EN2001a_H01_FEO066_0452218_0452528 And that is actually stored in another X. M. L. document.311AMI_EN2001a_H01_FEO066_0453820_0454533 Yeah, I w would then just not print out the um start and end times.312AMI_EN2001a_H01_FEO066_0455072_0455239 No, it's for every single word.313AMI_EN2001a_H01_FEO066_0455383_0455562 Or for every single utterance.314AMI_EN2001a_H01_FEO066_0455562_0455725 Yeah, that depends on what you want.315AMI_EN2001a_H01_FEO066_0455960_0456252 Yeah, but I do it with Perl, it's just string manipulation.316AMI_EN2001a_H01_FEO066_0456252_0456516 So I would I mean I would just317AMI_EN2001a_H01_FEO066_0456910_0457003 Sure.318AMI_EN2001a_H01_FEO066_0458242_0458429 No, I didn't do a sea no.319AMI_EN2001a_H01_FEO066_0460654_0461181 And you would want that all in one file for all the corpus?320AMI_EN2001a_H01_FEO066_0461181_0461204 Or321AMI_EN2001a_H01_FEO066_0461408_0461512 For the series.322AMI_EN2001a_H01_FEO066_0462018_0462314 Yeah, I can directly put it into uh just like323AMI_EN2001a_H01_FEO066_0462660_0463171 So uh only words um per meeting series.324AMI_EN2001a_H01_FEO066_0464375_0464413 Uh-huh.325AMI_EN2001a_H01_FEO066_0464813_0464889 Yes.326AMI_EN2001a_H01_FEO066_0465039_0465658 Yeah, they will just I will just take I would uh take over the names they have anyway.327AMI_EN2001a_H01_FEO066_0466108_0466147 Yeah, yeah.328AMI_EN2001a_H01_FEO066_0466281_0466707 Yeah, one series has the um same three starting letters.329AMI_EN2001a_H01_FEO066_0466707_0466734 So330AMI_EN2001a_H01_FEO066_0466910_0467212 So only words and words and times.331AMI_EN2001a_H01_FEO066_0467212_0467245 And you332AMI_EN2001a_H01_FEO066_0467792_0467946 Yeah, you want it ordered.333AMI_EN2001a_H01_FEO066_0467946_0468083 Okay.334AMI_EN2001a_H01_FEO066_0468083_0468316 Okay, anybody335AMI_EN2001a_H01_FEO066_0468537_0468688 Um ord base dot times.336AMI_EN2001a_H01_FEO066_0468810_0468925 Yeah, and do you want337AMI_EN2001a_H01_FEO066_0469097_0469418 Yeah, sometimes they're contained in one another.338AMI_EN2001a_H01_FEO066_0469418_0469458 So339AMI_EN2001a_H01_FEO066_0469642_0469880 Just after th mm-hmm.340AMI_EN2001a_H01_FEO066_0469880_0469910 'Kay.341AMI_EN2001a_H01_FEO066_0470034_0470162 Ordered.342AMI_EN2001a_H01_FEO066_0470162_0470257 Only words.343AMI_EN2001a_H01_FEO066_0472729_0473205 Um and I think um for all the corpus, it's just from I know from other times344AMI_EN2001a_H01_FEO066_0473205_0473933 it's um nine megami byte to have I mean should be should be similar to have the words.345AMI_EN2001a_H01_FEO066_0473933_0473987 Should be.346AMI_EN2001a_H01_FEO066_0474216_0474588 Na um all the words together um for all the meetings.347AMI_EN2001a_H01_FEO066_0474962_0475221 That's what I'm guessing that's, you know,348AMI_EN2001a_H01_FEO066_0475221_0476110 um what I because nine mega-byte is what I got for when I said for every um utterance, this is goes from there to there and349AMI_EN2001a_H01_FEO066_0476289_0476474 takes takes seconds.350AMI_EN2001a_H01_FEO066_0476785_0476897 Oh.351AMI_EN2001a_H01_FEO066_0476897_0477280 Yeah, I mean I'm it doing it for all of it.352AMI_EN2001a_H01_FEO066_0477280_0477364 Doesn't matter.353AMI_EN2001a_H01_FEO066_0479055_0479446 Yeah, I mean I hope it will be the same for the words.354AMI_EN2001a_H01_FEO066_0479446_0479529 It's just what I355AMI_EN2001a_H01_FEO066_0480946_0480952 I356AMI_EN2001a_H01_FEO066_0482485_0482572 Mm-hmm.357AMI_EN2001a_H01_FEO066_0482572_0482705 Mm.358AMI_EN2001a_H01_FEO066_0482705_0483935 So so um I will probably send um just one file of the first meeting um to all those who need it so that you can have a look whether that's what you want.359AMI_EN2001a_H01_FEO066_0484442_0484722 Yeah, I mean if it's just for one meeting, it's really not too big.360AMI_EN2001a_H01_FEO066_0484951_0484980 Yeah.361AMI_EN2001a_H01_FEO066_0504955_0505089 What do we have to demonstrate?362AMI_EN2001a_H02_FEO065_0021133_0021442 The basic word importance is off-line as well.363AMI_EN2001a_H02_FEO065_0021442_0022058 The combined measure might not be if we want to wait what the user has typed in into the search.364AMI_EN2001a_H02_FEO065_0070716_0070748 Yeah.365AMI_EN2001a_H02_FEO065_0081159_0081631 I'm not quite so what it did you want to do it, i you just wanted to assign366AMI_EN2001a_H02_FEO065_0082290_0082436 Uh I thought about words.367AMI_EN2001a_H02_FEO065_0083396_0083438 Mm.368AMI_EN2001a_H02_FEO065_0085553_0085612 Mm okay.369AMI_EN2001a_H02_FEO065_0089104_0089875 Yeah, but how about those words which don't carry any meaning at all, the um and uhs and something like that.370AMI_EN2001a_H02_FEO065_0090130_0090775 Because if we if we average average over over a whole utterance all the words,371AMI_EN2001a_H02_FEO065_0090775_0091688 and there are quite unimportant words in there, but quite important words as well, I think we should just disregard the the372AMI_EN2001a_H02_FEO065_0092528_0092595 Okay.373AMI_EN2001a_H02_FEO065_0093105_0093128 Alright.374AMI_EN2001a_H02_FEO065_0160400_0160526 Yeah.375AMI_EN2001a_H02_FEO065_0162071_0162292 But there is no I. D. for an utterance I think.376AMI_EN2001a_H02_FEO065_0162292_0162541 It's just for individual words.377AMI_EN2001a_H02_FEO065_0162541_0162802 So how do we do that then?378AMI_EN2001a_H02_FEO065_0162802_0163092 We for utterances as well.379AMI_EN2001a_H02_FEO065_0163384_0163615 I think it's just for one380AMI_EN2001a_H02_FEO065_0163876_0163922 word.381AMI_EN2001a_H02_FEO065_0163922_0164032 So we have to382AMI_EN2001a_H02_FEO065_0164327_0164362 Yeah.383AMI_EN2001a_H02_FEO065_0167471_0168000 Uh I'm not quite sure, I have only seen that the uh the individual words have got an I. D. .384AMI_EN2001a_H02_FEO065_0168332_0168396 Yeah.385AMI_EN2001a_H02_FEO065_0168396_0169241 You always could have a look at the time stamps and then take the ones that uh belong together to form an utterance.386AMI_EN2001a_H02_FEO065_0169821_0170181 Yeah, if they are already, there's it's easy but it would be possible.387AMI_EN2001a_H02_FEO065_0170334_0170490 Uh yeah.388AMI_EN2001a_H02_FEO065_0170490_0170523 Okay.389AMI_EN2001a_H02_FEO065_0223477_0224046 You s uh you said you are currently in uh implementing the idea.390AMI_EN2001a_H02_FEO065_0224046_0224300 What exactly are you computing?391AMI_EN2001a_H02_FEO065_0224466_0224526 Okay.392AMI_EN2001a_H02_FEO065_0224836_0224875 Okay.393AMI_EN2001a_H02_FEO065_0225136_0225169 Mm-hmm.394AMI_EN2001a_H02_FEO065_0243003_0243046 Mm-hmm.395AMI_EN2001a_H02_FEO065_0262785_0263972 Yeah, I w I w I would need the raw text pretty soon because I have to find out um how I have to put the segments into bins.396AMI_EN2001a_H02_FEO065_0263972_0264145 And then yeah.397AMI_EN2001a_H02_FEO065_0264441_0264633 No, that's not necessary.398AMI_EN2001a_H02_FEO065_0265498_0265629 Yes, I did.399AMI_EN2001a_H02_FEO065_0265629_0265910 But um I've only just got the notes.400AMI_EN2001a_H02_FEO065_0265910_0266503 I have to still have uh to order everything by the time and401AMI_EN2001a_H02_FEO065_0266714_0266935 Yeah, I think it's quite easy after the402AMI_EN2001a_H02_FEO065_0267137_0267219 Yeah.403AMI_EN2001a_H02_FEO065_0267219_0267267 Yeah.404AMI_EN2001a_H02_FEO065_0267376_0267433 So uh405AMI_EN2001a_H02_FEO065_0267873_0267898 Mm-hmm.406AMI_EN2001a_H02_FEO065_0268671_0268966 Yeah, b I uh w that's what I was uh thought.407AMI_EN2001a_H02_FEO065_0268966_0269391 That you just combine them and then order the time stamps accordingly.408AMI_EN2001a_H02_FEO065_0269991_0270115 Okay.409AMI_EN2001a_H02_FEO065_0270115_0270837 Um what I found out was that there are quite a lot of things without without s time stamps in the beginning.410AMI_EN2001a_H02_FEO065_0270998_0271224 Yeah, and uh X. M. L. files.411AMI_EN2001a_H02_FEO065_0271224_0271427 Yeah, that's just an I. D. or something.412AMI_EN2001a_H02_FEO065_0271427_0271464 I don't know.413AMI_EN2001a_H02_FEO065_0271464_0271554 Just numbers.414AMI_EN2001a_H02_FEO065_0271867_0272305 Yes, but what are the other things that's uh some kind of number?415AMI_EN2001a_H02_FEO065_0272305_0272671 F maybe the file number or something that is in the beginning.416AMI_EN2001a_H02_FEO065_0272671_0272729 What is that?417AMI_EN2001a_H02_FEO065_0272892_0272958 Do you know?418AMI_EN2001a_H02_FEO065_0273079_0273629 Um I think there are quite a lot of numbers in the beginning where n there is no time stamp for the numbers.419AMI_EN2001a_H02_FEO065_0273775_0273799 It's420AMI_EN2001a_H02_FEO065_0273989_0274476 Think they say um quite a lot of numbers and before that, uh um there's this number.421AMI_EN2001a_H02_FEO065_0274476_0274520 Was it422AMI_EN2001a_H02_FEO065_0274728_0275313 Yeah, there i are numbers in the um the W. tag, but there are no time stamps.423AMI_EN2001a_H02_FEO065_0275688_0275740 Yeah.424AMI_EN2001a_H02_FEO065_0276303_0276610 Yeah, in the beginning as well sometimes, I think.425AMI_EN2001a_H02_FEO065_0276610_0276756 At least I saw some.426AMI_EN2001a_H02_FEO065_0276933_0276970 Yeah.427AMI_EN2001a_H02_FEO065_0277529_0277556 Yeah.428AMI_EN2001a_H02_FEO065_0280836_0281081 But what it is it actually that numbers?429AMI_EN2001a_H02_FEO065_0281341_0281756 Okay, so but there are no time stamps annotated to that.430AMI_EN2001a_H02_FEO065_0281756_0281882 It's it's quite strange.431AMI_EN2001a_H02_FEO065_0284445_0285042 And also um there are different um combinations of letters.432AMI_EN2001a_H02_FEO065_0285042_0285231 B. R. E. and something like that.433AMI_EN2001a_H02_FEO065_0285231_0285858 Is it everything ordered are the time stamps global or uh are they local at any point?434AMI_EN2001a_H02_FEO065_0286265_0286303 Okay.435AMI_EN2001a_H02_FEO065_0291823_0291951 Yeah, it's Rainbow.436AMI_EN2001a_H02_FEO065_0291951_0292365 It's um I think it's just the dictionary in the first place.437AMI_EN2001a_H02_FEO065_0292365_0292385 But438AMI_EN2001a_H02_FEO065_0293589_0294373 Um no, I have to bin it up and so I will only have counts for each each bin or something.439AMI_EN2001a_H02_FEO065_0294773_0295105 It's because um Rainbow is a text classification system.440AMI_EN2001a_H02_FEO065_0295105_0295659 And I think it's not possible to have just one class.441AMI_EN2001a_H02_FEO065_0295659_0295760 That's the problem.442AMI_EN2001a_H02_FEO065_0296125_0296249 Maybe we could443AMI_EN2001a_H02_FEO065_0296580_0296937 Yeah sure, you sure, we could do that, but I don't that makes sense.444AMI_EN2001a_H02_FEO065_0297812_0298677 If we need just frequencies, maybe we should just calculate them by using Perl or something.445AMI_EN2001a_H02_FEO065_0298677_0298718 I don't know.446AMI_EN2001a_H02_FEO065_0299138_0299656 Yeah, it's quite easy to just count and s or sort them by um frequency.447AMI_EN2001a_H02_FEO065_0299861_0300069 Just using a Perl script.448AMI_EN2001a_H02_FEO065_0300549_0300625 Is it too big?449AMI_EN2001a_H02_FEO065_0301052_0301110 Yeah.450AMI_EN2001a_H02_FEO065_0301958_0302002 Hmm.451AMI_EN2001a_H02_FEO065_0302614_0303046 I don't know how you how many terms you can handle in Perl.452AMI_EN2001a_H02_FEO065_0303774_0303808 Mm yeah.453AMI_EN2001a_H02_FEO065_0349364_0349721 Uh I can get all the raw text, but it has to be ordered still.454AMI_EN2001a_H02_FEO065_0349721_0349749 So455AMI_EN2001a_H02_FEO065_0350232_0350297 No, it isn't.456AMI_EN2001a_H02_FEO065_0352339_0353097 Um it's in what is implemented in Rainbow is information gain, and I'm not quite sure how they calculate that.457AMI_EN2001a_H02_FEO065_0353720_0353762 Yeah.458AMI_EN2001a_H02_FEO065_0356340_0356461 Uh that's what Rainbow does.459AMI_EN2001a_H02_FEO065_0356461_0357077 I think you j can just get probabilities for a certain words for each document.460AMI_EN2001a_H02_FEO065_0357206_0357252 Certain461AMI_EN2001a_H02_FEO065_0357535_0357890 Um we would have to look at that.462AMI_EN2001a_H02_FEO065_0358558_0358602 Mm-hmm.463AMI_EN2001a_H02_FEO065_0358779_0358810 Oh.464AMI_EN2001a_H02_FEO065_0378348_0379352 Yeah, that's what I thought as well, that you that probably the the topic segment level is the most um informative for the words.465AMI_EN2001a_H02_FEO065_0379630_0379762 Yeah, that's the problem.466AMI_EN2001a_H02_FEO065_0379762_0379823 I don't know.467AMI_EN2001a_H02_FEO065_0414949_0414996 Mm-hmm.468AMI_EN2001a_H02_FEO065_0425222_0425435 So shall we sit together tomorrow then as well?469AMI_EN2001a_H02_FEO065_0425435_0425576 Uh Okay.470AMI_EN2001a_H02_FEO065_0426508_0426760 Um, yeah, w would it be best?471AMI_EN2001a_H02_FEO065_0426760_0427074 At the moment it's it's just lines of472AMI_EN2001a_H02_FEO065_0427303_0427339 Mm-hmm.473AMI_EN2001a_H02_FEO065_0429613_0429670 Um474AMI_EN2001a_H02_FEO065_0430307_0430360 Okay.475AMI_EN2001a_H02_FEO065_0430360_0430795 So um you'd do you extract the words, the raw text, as well?476AMI_EN2001a_H02_FEO065_0430795_0430822 Uh477AMI_EN2001a_H02_FEO065_0431250_0431281 Okay.478AMI_EN2001a_H02_FEO065_0431522_0431557 Mm-hmm.479AMI_EN2001a_H02_FEO065_0431796_0431853 Print out.480AMI_EN2001a_H02_FEO065_0433668_0433734 Okay.481AMI_EN2001a_H02_FEO065_0433734_0433787 Okay, that482AMI_EN2001a_H02_FEO065_0434280_0434317 Okay.483AMI_EN2001a_H02_FEO065_0435134_0435382 So have we already extracted from all the files?484AMI_EN2001a_H02_FEO065_0435606_0435639 Yeah.485AMI_EN2001a_H02_FEO065_0435874_0436021 Did you also order486AMI_EN2001a_H02_FEO065_0436159_0436181 Mm-hmm.487AMI_EN2001a_H02_FEO065_0436920_0436957 Hmm.488AMI_EN2001a_H02_FEO065_0441180_0441207 Hmm.489AMI_EN2001a_H02_FEO065_0454555_0454593 Okay.490AMI_EN2001a_H02_FEO065_0467222_0467546 Uh I don't need the times, I just need the words.491AMI_EN2001a_H02_FEO065_0467546_0467696 But um492AMI_EN2001a_H02_FEO065_0467818_0467918 Yeah, in the right order.493AMI_EN2001a_H02_FEO065_0467918_0468245 Yes.494AMI_EN2001a_H02_FEO065_0469487_0469710 Yeah, that doesn't matter too much, I think.495AMI_EN2001a_H02_FEO065_0482534_0482559 Hmm.496AMI_EN2001a_H02_FEO065_0483991_0484027 Mm-hmm.497AMI_EN2001a_H02_FEO065_0485657_0486309 How long would it take to make the frequency counts with a Java hash table?498AMI_EN2001a_H02_FEO065_0486820_0486882 Yeah.499AMI_EN2001a_H02_FEO065_0487104_0487397 No, how long you would have to program something.500AMI_EN2001a_H02_FEO065_0487980_0488010 Okay.501AMI_EN2001a_H02_FEO065_0488637_0488672 Mm.502AMI_EN2001a_H02_FEO065_0489118_0489972 Because it's quite easy in Perl as well, it's just a line of code for counting all the words and yeah, it's it's by hashes.503AMI_EN2001a_H02_FEO065_0490168_0490205 Yeah.504AMI_EN2001a_H02_FEO065_0490503_0490527 Yeah.505AMI_EN2001a_H02_FEO065_0491512_0491539 'Kay.506AMI_EN2001a_H03_MEE067_0001707_0001961 I I dry-read it the last time.507AMI_EN2001a_H03_MEE067_0003878_0003940 Next week.508AMI_EN2001a_H03_MEE067_0009375_0009423 Yeah.509AMI_EN2001a_H03_MEE067_0010977_0011022 Yeah.510AMI_EN2001a_H03_MEE067_0016034_0016072 No.511AMI_EN2001a_H03_MEE067_0022211_0022538 Uh mine's gonna be mostly using the off-line.512AMI_EN2001a_H03_MEE067_0022538_0022728 But the actual stuff it's doing will be on-line.513AMI_EN2001a_H03_MEE067_0022728_0023221 But it won't be very um processor intensive or memory intensive, I don't think.514AMI_EN2001a_H03_MEE067_0029486_0029552 Don't think so.515AMI_EN2001a_H03_MEE067_0029733_0029764 Yeah.516AMI_EN2001a_H03_MEE067_0031094_0031358 Are we still gonna go for dumping it into a database?517AMI_EN2001a_H03_MEE067_0031358_0031560 Are we still gonna dump it into a database?518AMI_EN2001a_H03_MEE067_0032020_0032405 'Cause if we are, I reckon we should all read our classes out of the database.519AMI_EN2001a_H03_MEE067_0032405_0032529 It'll be so much easier.520AMI_EN2001a_H03_MEE067_0032669_0033015 Well if we're gonna dump the part of it into a database anyway,521AMI_EN2001a_H03_MEE067_0033015_0033618 we might as well dump all the fields we want into the database, calculate everything from there.522AMI_EN2001a_H03_MEE067_0033618_0034110 Then we don't even have to worry that much about the underlying X. M. L. representation.523AMI_EN2001a_H03_MEE067_0034110_0034202 We can just query it.524AMI_EN2001a_H03_MEE067_0039366_0039887 Well if we're gonna do that, we should try and store everything in in an X. M. L. format as well.525AMI_EN2001a_H03_MEE067_0040119_0040152 Yeah.526AMI_EN2001a_H03_MEE067_0040825_0040856 Yeah.527AMI_EN2001a_H03_MEE067_0049453_0049889 Well we don't even need to do that, 'cause if we got our information density calculated off-line528AMI_EN2001a_H03_MEE067_0049889_0050247 so all we do is treat the whole lot as one massive document.529AMI_EN2001a_H03_MEE067_0050247_0050912 I mean they'll it's not gonna be so big that we can't load in a information density for every utterance.530AMI_EN2001a_H03_MEE067_0050912_0051079 And we can just summarise based on that.531AMI_EN2001a_H03_MEE067_0056786_0056925 I think you can do it on-line.532AMI_EN2001a_H03_MEE067_0057891_0058716 I don't think there's really much point in doing like that when it's just gonna feed off in the end the information density measure basically.533AMI_EN2001a_H03_MEE067_0058716_0058885 And that's all calculated off-line.534AMI_EN2001a_H03_MEE067_0058885_0059458 So what you're really doing is sorting a list, is the p computationally hard part of it.535AMI_EN2001a_H03_MEE067_0059884_0060690 Well like the ideas we're calculating are information density all off-line first for every utterance in the whole corpus, right?536AMI_EN2001a_H03_MEE067_0060690_0061497 So what you do is you say if you're looking at a series of meetings, you just say well our whole document comprises of all these stuck together.537AMI_EN2001a_H03_MEE067_0061497_0062017 And then all you have to do is sort them by j information density.538AMI_EN2001a_H03_MEE067_0062017_0062382 Like maybe weighted with the search terms, and then extract them.539AMI_EN2001a_H03_MEE067_0062382_0062746 I don't think it's too slow to do on-line, to be honest.540AMI_EN2001a_H03_MEE067_0063926_0063965 Is that541AMI_EN2001a_H03_MEE067_0065445_0065476 Yeah.542AMI_EN2001a_H03_MEE067_0068569_0068829 Well, on the utterance level I was thinking.543AMI_EN2001a_H03_MEE067_0068829_0069254 So the utterances with the highest like mean information density.544AMI_EN2001a_H03_MEE067_0069570_0069883 Well the trouble with doing it on the word level is if you want the audio to synch up545AMI_EN2001a_H03_MEE067_0069883_0070238 you've got no way of getting in and extracting just that word.546AMI_EN2001a_H03_MEE067_0070238_0070325 I mean it's impossible.547AMI_EN2001a_H03_MEE067_0070589_0070782 For every single word?548AMI_EN2001a_H03_MEE067_0070782_0070838 Oh, okay.549AMI_EN2001a_H03_MEE067_0071024_0071050 Yeah.550AMI_EN2001a_H03_MEE067_0071395_0071549 I don't think that will do it.551AMI_EN2001a_H03_MEE067_0071549_0071640 We'll have to buffer it.552AMI_EN2001a_H03_MEE067_0073408_0073626 Well the skimming's gonna use the importance.553AMI_EN2001a_H03_MEE067_0073845_0074084 But like at first it's just gonna be I. D. F. .554AMI_EN2001a_H03_MEE067_0074502_0074630 Well mostly skimming, yeah.555AMI_EN2001a_H03_MEE067_0077743_0077774 Yeah.556AMI_EN2001a_H03_MEE067_0079221_0079563 Well the nice thing about that is it will automatically be in sentences.557AMI_EN2001a_H03_MEE067_0079563_0079785 Well more or less.558AMI_EN2001a_H03_MEE067_0079785_0080046 So it will make more sense, and if you get just extract words.559AMI_EN2001a_H03_MEE067_0084650_0084684 Yeah.560AMI_EN2001a_H03_MEE067_0087034_0087088 I see it.561AMI_EN2001a_H03_MEE067_0087250_0088042 But it'll need to be calculated at word level though because otherwise there won't be enough occurrences of the terms to make any meaningful sense.562AMI_EN2001a_H03_MEE067_0088415_0088444 Yeah.563AMI_EN2001a_H03_MEE067_0088974_0089228 Yeah, I reckon you can just mean it over the sentence.564AMI_EN2001a_H03_MEE067_0090192_0090335 I think we should filter them.565AMI_EN2001a_H03_MEE067_0091714_0091999 Maybe we should have like um a cut-off.566AMI_EN2001a_H03_MEE067_0091999_0092527 So it a w word only gets a value if it's above a certain threshold.567AMI_EN2001a_H03_MEE067_0092527_0093049 So anything that has less than say nought point five importance gets assigned to zero.568AMI_EN2001a_H03_MEE067_0093333_0093440 Yeah, that's the other th569AMI_EN2001a_H03_MEE067_0093947_0093984 Yeah.570AMI_EN2001a_H03_MEE067_0095821_0095998 I think we'll have to buffer the audio.571AMI_EN2001a_H03_MEE067_0095998_0096167 But I don't think it will be very hard.572AMI_EN2001a_H03_MEE067_0096167_0096348 I think it would be like an hour or two's work.573AMI_EN2001a_H03_MEE067_0096630_0097045 Like just build an another f wave file essentially.574AMI_EN2001a_H03_MEE067_0097179_0097414 Yeah, I mean I bet there would be packages575AMI_EN2001a_H03_MEE067_0097639_0097766 In memory, yeah.576AMI_EN2001a_H03_MEE067_0097766_0098250 So just like unp there's bound to be like a media wave object or something like that.577AMI_EN2001a_H03_MEE067_0098250_0098399 And just build one in memory.578AMI_EN2001a_H03_MEE067_0098731_0098777 I don't know.579AMI_EN2001a_H03_MEE067_0098777_0098944 I have no idea.580AMI_EN2001a_H03_MEE067_0098944_0099297 But it must have like classes for dealing with files.581AMI_EN2001a_H03_MEE067_0099297_0099721 And if it has classes for concatenating files, you can do it in memory.582AMI_EN2001a_H03_MEE067_0099721_0099740 So583AMI_EN2001a_H03_MEE067_0101660_0102614 Well what I think I might try and build is basically a class that you just feed it a linked list of um different wave-forms,584AMI_EN2001a_H03_MEE067_0102614_0103261 and it will just string them all together with maybe, I don't know, tenth of a second silence in between each one or something like that.585AMI_EN2001a_H03_MEE067_0104116_0104228 Normalise it, yeah.586AMI_EN2001a_H03_MEE067_0107283_0107415 Oh yeah, yeah, we'll need that.587AMI_EN2001a_H03_MEE067_0107621_0107910 We also really wanna be able to search by who's speaking as well.588AMI_EN2001a_H03_MEE067_0109242_0109529 It doesn't matter, 'cause all the calculation's done off-line.589AMI_EN2001a_H03_MEE067_0127833_0127892 That's easy.590AMI_EN2001a_H03_MEE067_0127892_0128249 You just like create a new X. M. L. document in memory.591AMI_EN2001a_H03_MEE067_0133618_0134351 I don't think it's really that much of a problem because if it's too big, what we can do is just well all the off-line stuff doesn't really matter.592AMI_EN2001a_H03_MEE067_0134351_0134673 And all we can do is just process a bit at a time.593AMI_EN2001a_H03_MEE067_0134673_0135532 Like for summarisation, say we wanted a hundred utterances in the summary, just look at the meeting, take the top one hundred utterances in each other meeting.594AMI_EN2001a_H03_MEE067_0135532_0136099 If it scores higher than the ones already in the summary so far, just replace them.595AMI_EN2001a_H03_MEE067_0136099_0136309 And then you only have to process one meeting at a time.596AMI_EN2001a_H03_MEE067_0138344_0139072 Okay, so maybe we should build a b store a mean measure for the segments and meetings as well?597AMI_EN2001a_H03_MEE067_0142298_0142500 And speaker.598AMI_EN2001a_H03_MEE067_0142500_0143072 Speaker and um topic segmenting we'll need as well.599AMI_EN2001a_H03_MEE067_0143289_0143322 Yeah.600AMI_EN2001a_H03_MEE067_0144840_0145325 Well yeah, and then it'll f preserve the order when it's displayed the601AMI_EN2001a_H03_MEE067_0145483_0145500 Yeah.602AMI_EN2001a_H03_MEE067_0145708_0145775 Yeah.603AMI_EN2001a_H03_MEE067_0146626_0146711 Yeah, I think so.604AMI_EN2001a_H03_MEE067_0155756_0156186 So we should basically make our own X. M. L. document in memory605AMI_EN2001a_H03_MEE067_0156392_0156585 that everyone's um606AMI_EN2001a_H03_MEE067_0156730_0157162 module changes that, rather than the underlying data.607AMI_EN2001a_H03_MEE067_0157162_0157650 And then have that X. M. L. uh NITE X. M. L. document tied to the interface.608AMI_EN2001a_H03_MEE067_0157986_0158169 Well, you can make it in a file if you want.609AMI_EN2001a_H03_MEE067_0160340_0160372 Mm-hmm.610AMI_EN2001a_H03_MEE067_0174275_0174475 They are utterances, aren't they?611AMI_EN2001a_H03_MEE067_0174475_0174754 The segments are utterances, aren't they?612AMI_EN2001a_H03_MEE067_0174754_0174777 Yeah.613AMI_EN2001a_H03_MEE067_0175055_0175131 Alright, okay.614AMI_EN2001a_H03_MEE067_0176892_0176988 Well, that's easy.615AMI_EN2001a_H03_MEE067_0178952_0179132 Well it's close enough, isn't it?616AMI_EN2001a_H03_MEE067_0179132_0179522 It may not be exact every time, but it's a so sort of size we're looking for.617AMI_EN2001a_H03_MEE067_0183632_0183699 Yeah, yeah.618AMI_EN2001a_H03_MEE067_0192349_0192375 Yeah.619AMI_EN2001a_H03_MEE067_0193162_0193473 But why don't we just write it as a new X. M. L. file?620AMI_EN2001a_H03_MEE067_0193473_0194041 Can NITE handle just loading arbitrary uh new like attributes and stuff?621AMI_EN2001a_H03_MEE067_0194175_0194354 I mean, I would have thought they'd make it able to.622AMI_EN2001a_H03_MEE067_0194498_0194527 Yeah.623AMI_EN2001a_H03_MEE067_0194919_0195221 So why do we need to have two X. M. L. trees in memory at once?624AMI_EN2001a_H03_MEE067_0198046_0198680 The other thing is that would mean we'd be using their parser as well, which means we wouldn't have to parse anything, which be quite nice.625AMI_EN2001a_H03_MEE067_0198680_0199001 'Cause their parser is probably much faster than anything we've come up with anyway.626AMI_EN2001a_H03_MEE067_0205106_0205459 Yeah, I mean we can process it in chunks if it gets too big basically.627AMI_EN2001a_H03_MEE067_0205459_0205894 We can just process it all in chunks if it gets too big to load it into memory.628AMI_EN2001a_H03_MEE067_0208298_0208616 I think we probably want to store Sorry.629AMI_EN2001a_H03_MEE067_0208616_0209230 I think we probably want to store um a hierarchical information density as well.630AMI_EN2001a_H03_MEE067_0209230_0209819 So like an informan mation density score for each meeting and each topic segment.631AMI_EN2001a_H03_MEE067_0209819_0210144 'Cause otherwise we'd be recalculating the same thing over and over and over again.632AMI_EN2001a_H03_MEE067_0210288_0210321 Yeah.633AMI_EN2001a_H03_MEE067_0210498_0210848 And that will obviously make it much easier to display.634AMI_EN2001a_H03_MEE067_0211144_0211355 Well it may not for the whole meeting, but like635AMI_EN2001a_H03_MEE067_0212594_0212708 Yeah, exactly.636AMI_EN2001a_H03_MEE067_0213042_0213069 Yeah.637AMI_EN2001a_H03_MEE067_0214453_0214567 Well, we can start off638AMI_EN2001a_H03_MEE067_0214769_0214815 like that.639AMI_EN2001a_H03_MEE067_0214815_0215346 Well I was gonna start off I've v got sort of half-way through implementing one that does just I. D. F. .640AMI_EN2001a_H03_MEE067_0215509_0215863 And then just I can change that to work on whatever.641AMI_EN2001a_H03_MEE067_0216954_0216988 Yeah.642AMI_EN2001a_H03_MEE067_0217138_0217756 And it should be weighted by stuff like the hot spots and um the key-words in the search and stuff like that.643AMI_EN2001a_H03_MEE067_0220760_0220958 Did he not say something about named entities?644AMI_EN2001a_H03_MEE067_0220958_0221163 So I thought he said there wasn't very many.645AMI_EN2001a_H03_MEE067_0221300_0221344 Yeah.646AMI_EN2001a_H03_MEE067_0224048_0224194 Yeah.647AMI_EN2001a_H03_MEE067_0224194_0224651 It's not T. F. I. D. F., it's just inverse document frequency.648AMI_EN2001a_H03_MEE067_0224651_0224942 'Cause it's really easy to do basically.649AMI_EN2001a_H03_MEE067_0224942_0225115 There's just like for a baseline really.650AMI_EN2001a_H03_MEE067_0225442_0225567 Well, I'm half-way through.651AMI_EN2001a_H03_MEE067_0225567_0225761 It's not working yet, but it will do.652AMI_EN2001a_H03_MEE067_0226250_0226429 Um yeah.653AMI_EN2001a_H03_MEE067_0226429_0226640 And then averaging it over the utterances.654AMI_EN2001a_H03_MEE067_0226785_0227073 But it's not like um related to the corpus at all.655AMI_EN2001a_H03_MEE067_0227073_0227435 It's just working on an arbitrary text file at the moment.656AMI_EN2001a_H03_MEE067_0227757_0227786 No.657AMI_EN2001a_H03_MEE067_0230509_0230969 It would be useful to know how everyone's gonna store their things though.658AMI_EN2001a_H03_MEE067_0231551_0231588 Yeah.659AMI_EN2001a_H03_MEE067_0232466_0232562 Yeah.660AMI_EN2001a_H03_MEE067_0232562_0232846 Well I've got like a few hours free.661AMI_EN2001a_H03_MEE067_0232846_0232936 Like after this.662AMI_EN2001a_H03_MEE067_0242869_0243009 It's the most boring task.663AMI_EN2001a_H03_MEE067_0243171_0243217 Yeah.664AMI_EN2001a_H03_MEE067_0245429_0245727 Or at least um simple versions of them.665AMI_EN2001a_H03_MEE067_0250500_0250962 So maybe we should try doing something really simple, like just displaying a whole meeting.666AMI_EN2001a_H03_MEE067_0251087_0251385 And like just being able to scroll through it or something like that.667AMI_EN2001a_H03_MEE067_0251604_0251651 Yeah.668AMI_EN2001a_H03_MEE067_0253781_0253904 Are you free after this?669AMI_EN2001a_H03_MEE067_0254605_0254723 How about Friday then.670AMI_EN2001a_H03_MEE067_0254723_0254835 'Cause I'm off all Friday.671AMI_EN2001a_H03_MEE067_0255266_0255551 Uh Wednesday I've got a nine 'til twelve.672AMI_EN2001a_H03_MEE067_0256121_0256321 Yeah, nothing in the afternoon.673AMI_EN2001a_H03_MEE067_0256321_0256473 I've got nothing in the afternoon.674AMI_EN2001a_H03_MEE067_0256473_0256489 So675AMI_EN2001a_H03_MEE067_0256655_0256719 Okay.676AMI_EN2001a_H03_MEE067_0256719_0256803 So you ha yeah.677AMI_EN2001a_H03_MEE067_0257084_0257228 Where about, just in Appleton Tower?678AMI_EN2001a_H03_MEE067_0257671_0258002 Uh I'll be in um the Appleton Tower anyway.679AMI_EN2001a_H03_MEE067_0258234_0258565 Um well I'll be there from twelve.680AMI_EN2001a_H03_MEE067_0258565_0259130 I've got some other stuff that needs done on Matlab, so if you're not there at twelve, I can just work on that.681AMI_EN2001a_H03_MEE067_0259130_0259144 So682AMI_EN2001a_H03_MEE067_0259287_0259329 Yeah.683AMI_EN2001a_H03_MEE067_0265027_0265071 Why w684AMI_EN2001a_H03_MEE067_0288510_0288580 Yeah.685AMI_EN2001a_H03_MEE067_0288580_0288724 I'm just building a dictionary.686AMI_EN2001a_H03_MEE067_0289034_0289496 Oh, mine's just gonna use the um hash map one in um Java.687AMI_EN2001a_H03_MEE067_0289496_0289725 'Cause I'm only gonna do it on small documents.688AMI_EN2001a_H03_MEE067_0289725_0290180 It's just like bef until the information density is up and running.689AMI_EN2001a_H03_MEE067_0290180_0290368 Just something to get give me something to work with.690AMI_EN2001a_H03_MEE067_0290584_0290903 So it's only gonna use quite small documents, you see, to start with.691AMI_EN2001a_H03_MEE067_0294403_0294736 Why does it need to be classified into like different segments?692AMI_EN2001a_H03_MEE067_0295760_0296113 Can we just fill a second class with junk that we don't care about?693AMI_EN2001a_H03_MEE067_0296353_0296603 Like, I don't know, copies of Shakespeare or something.694AMI_EN2001a_H03_MEE067_0296955_0297545 'Cause if what we're looking for is the um frequency statistics, I don't see how that would be changed by the classification.695AMI_EN2001a_H03_MEE067_0298486_0298558 I the696AMI_EN2001a_H03_MEE067_0298828_0299012 Well there maybe another tool available?697AMI_EN2001a_H03_MEE067_0299444_0299475 Yeah.698AMI_EN2001a_H03_MEE067_0304560_0304830 Um I can't remember who's got it.699AMI_EN2001a_H03_MEE067_0304830_0304947 Might be WordNet.700AMI_EN2001a_H03_MEE067_0304947_0306137 But one of these big corpuses has a list of stop words that you can download and they're just basically lists of really uninteresting boring words that we could filter out before we do that.701AMI_EN2001a_H03_MEE067_0306457_0306661 It's like that's one the papers I read702AMI_EN2001a_H03_MEE067_0306661_0307513 that's um one things they did right at the beginning is they've got this big s stop-list and they just ignore all of those throughout the experiment.703AMI_EN2001a_H03_MEE067_0309109_0309372 Yeah, I it would be useful for me as well.704AMI_EN2001a_H03_MEE067_0309710_0310059 It uh I think that'd be useful for me as well.705AMI_EN2001a_H03_MEE067_0310059_0310095 Yeah.706AMI_EN2001a_H03_MEE067_0310595_0310631 Yeah.707AMI_EN2001a_H03_MEE067_0311547_0312168 Well all you really wanna do is look into getting some sub-set of the ICSI corpus off the DICE machines.708AMI_EN2001a_H03_MEE067_0312168_0312309 'Cause I hate working on DICE.709AMI_EN2001a_H03_MEE067_0312309_0312379 It's awful.710AMI_EN2001a_H03_MEE067_0312551_0312728 Like so I can use my home machine.711AMI_EN2001a_H03_MEE067_0313000_0313283 ha has a C. D. burner though.712AMI_EN2001a_H03_MEE067_0313283_0313399 has a C. D. burner.713AMI_EN2001a_H03_MEE067_0314936_0315052 Yeah.714AMI_EN2001a_H03_MEE067_0315052_0315220 The right-hand corner, far right.715AMI_EN2001a_H03_MEE067_0315430_0315462 Yeah.716AMI_EN2001a_H03_MEE067_0317027_0317238 How big is it without um717AMI_EN2001a_H03_MEE067_0317452_0317670 the WAV files and stuff?718AMI_EN2001a_H03_MEE067_0317812_0318369 'Cause I could just say at um going over S. C. P. one night and just leave it going all night if I had to.719AMI_EN2001a_H03_MEE067_0319290_0319815 It's yeah, I mean the wave data are obviously not gonna get off there completely.720AMI_EN2001a_H03_MEE067_0320583_0320670 Really?721AMI_EN2001a_H03_MEE067_0320670_0320720 Oh right?722AMI_EN2001a_H03_MEE067_0321476_0321705 I'll see if I can S. C. P. it, I suppose.723AMI_EN2001a_H03_MEE067_0322097_0322345 I've got a Linux box and a Windows box.724AMI_EN2001a_H03_MEE067_0322345_0322386 So725AMI_EN2001a_H03_MEE067_0322556_0322604 Broad-band.726AMI_EN2001a_H03_MEE067_0322932_0323099 Put it on to C. D. .727AMI_EN2001a_H03_MEE067_0323099_0323310 I can if I get down I can put to C. D. .728AMI_EN2001a_H03_MEE067_0323471_0323505 Yeah.729AMI_EN2001a_H03_MEE067_0324717_0324944 I'm not sure if there's enough space.730AMI_EN2001a_H03_MEE067_0324944_0325090 Is how much do we get?731AMI_EN2001a_H03_MEE067_0325402_0325496 Really?732AMI_EN2001a_H03_MEE067_0325496_0325559 Okay.733AMI_EN2001a_H03_MEE067_0327248_0327483 Yeah, but I can do it from that session, can't I?734AMI_EN2001a_H03_MEE067_0327483_0327943 You can compress it from a remote session and S. C. P. it from the same session?735AMI_EN2001a_H03_MEE067_0330026_0330074 Do you think?736AMI_EN2001a_H03_MEE067_0330585_0330689 Yeah.737AMI_EN2001a_H03_MEE067_0330689_0331280 Oh no no, I was thinking of SSHing just into some machine and then just SCPing it from there.738AMI_EN2001a_H03_MEE067_0331713_0331751 Yeah.739AMI_EN2001a_H03_MEE067_0331751_0331903 I mean it has to go through the gateway.740AMI_EN2001a_H03_MEE067_0331903_0331917 But741AMI_EN2001a_H03_MEE067_0332134_0332220 Can you not do that?742AMI_EN2001a_H03_MEE067_0332832_0332894 Mm, I see.743AMI_EN2001a_H03_MEE067_0333622_0333656 Yeah.744AMI_EN2001a_H03_MEE067_0341914_0341971 So you could just745AMI_EN2001a_H03_MEE067_0342503_0342756 But th first, uh how big are the chunks?746AMI_EN2001a_H03_MEE067_0342917_0343095 How big are the chunks you're looking at?747AMI_EN2001a_H03_MEE067_0344213_0344421 So quite small then.748AMI_EN2001a_H03_MEE067_0344421_0344970 So you could just um you could use just the same thing we used to build the big dictionary.749AMI_EN2001a_H03_MEE067_0344970_0345708 You just do that on-line 'cause that won't take long to build a little dictionary that big, will it.750AMI_EN2001a_H03_MEE067_0345708_0346083 I mean just use the same tool that we use.751AMI_EN2001a_H03_MEE067_0346083_0346105 Yeah.752AMI_EN2001a_H03_MEE067_0346487_0346521 Yeah.753AMI_EN2001a_H03_MEE067_0350090_0350212 It doesn't need ordered, no.754AMI_EN2001a_H03_MEE067_0351667_0352274 Um well that's the t are you using T. F. I. D. F. for the information density?755AMI_EN2001a_H03_MEE067_0352845_0352939 Alright, okay.756AMI_EN2001a_H03_MEE067_0353242_0353770 Like 'cause frequency would be useful, I think.757AMI_EN2001a_H03_MEE067_0353770_0354792 But um depending on the context, the size, and what we consider a document in the sense of calculating T. F. I. D. F. is gonna change.758AMI_EN2001a_H03_MEE067_0355046_0355219 Which might need thinking about.759AMI_EN2001a_H03_MEE067_0359903_0360060 I think it would be useful, yeah.760AMI_EN2001a_H03_MEE067_0360239_0360740 Well you need the raw frequency as well.761AMI_EN2001a_H03_MEE067_0360740_0360805 But um762AMI_EN2001a_H03_MEE067_0360952_0361581 you also need how many times things occur within each document.763AMI_EN2001a_H03_MEE067_0361734_0362273 And um what we consider a document's gonna depend on our context, I think.764AMI_EN2001a_H03_MEE067_0362273_0362939 'Cause if we're looking at the whole lot of meetings, we'll consider each meeting a document in sort of terms of this algorithm.765AMI_EN2001a_H03_MEE067_0362939_0363862 And if we're viewing like say just a small topic segment you might look at even each utterance as a small document.766AMI_EN2001a_H03_MEE067_0365712_0365859 Yeah, but the thing is um767AMI_EN2001a_H03_MEE067_0366204_0366444 It's gonna need some th th thought of how we768AMI_EN2001a_H03_MEE067_0366617_0366826 Actually maybe it doesn't actually matter.769AMI_EN2001a_H03_MEE067_0366826_0367304 Maybe if you just do it once at the highest level, it it will be fine.770AMI_EN2001a_H03_MEE067_0367304_0368111 But I was just thinking it might be difficult to calculate the T. F. I. D. F. off-line for all the different levels we might want.771AMI_EN2001a_H03_MEE067_0368111_0368920 'Cause if we're gonna allow disjoint segments for example, then how are we gonna know what's gonna be in context at any given time?772AMI_EN2001a_H03_MEE067_0368920_0369395 But I suppose if you just did it globally, treating a meeting as a document,773AMI_EN2001a_H03_MEE067_0369395_0370025 it'd probably still be work out fine, because you'd only be comparing to ones within the context.774AMI_EN2001a_H03_MEE067_0370973_0371258 Uh I don't know, I thought were you gonna use that in the end?775AMI_EN2001a_H03_MEE067_0371540_0371671 The information density.776AMI_EN2001a_H03_MEE067_0372940_0373031 Oh sorry, that's what I mean.777AMI_EN2001a_H03_MEE067_0373031_0373731 Like um yeah, for each word or whatever, but across the whole lot is what I mean by highest level.778AMI_EN2001a_H03_MEE067_0373731_0373857 Like across the whole corpus.779AMI_EN2001a_H03_MEE067_0374673_0374929 Yeah, but you'd probably look at each meeting as a document.780AMI_EN2001a_H03_MEE067_0376642_0376734 Mm possibly.781AMI_EN2001a_H03_MEE067_0379442_0379705 Are they big enough to get anything meaningful out of?782AMI_EN2001a_H03_MEE067_0381421_0381578 Well yeah, that is not it's not an issue.783AMI_EN2001a_H03_MEE067_0381578_0381827 You just concatenate an X. M. L. file together.784AMI_EN2001a_H03_MEE067_0381827_0382265 but we still want to have like a notion of meetings for the user.785AMI_EN2001a_H03_MEE067_0383098_0383210 Yeah, sure.786AMI_EN2001a_H03_MEE067_0383210_0383948 Yeah, you just like whatever you want to look at, you just jam together into an X. M. L. file and that's your meeting787AMI_EN2001a_H03_MEE067_0383948_0384337 even though bits of it may come from all over the place or whatever.788AMI_EN2001a_H03_MEE067_0384337_0384803 I mean I don't see why that's really a big problem.789AMI_EN2001a_H03_MEE067_0385698_0386254 So basically what you're saying is you can take an arbitrary amount of data and process it with the same algorithm.790AMI_EN2001a_H03_MEE067_0386254_0386626 It doesn't matter conceptually what that data is.791AMI_EN2001a_H03_MEE067_0386626_0386737 It could be a meeting.792AMI_EN2001a_H03_MEE067_0386737_0386927 it could be two utterances.793AMI_EN2001a_H03_MEE067_0386927_0387216 it could be a meeting plus half a meeting from somewhere else.794AMI_EN2001a_H03_MEE067_0390471_0390610 I don't think it's very difficult though.795AMI_EN2001a_H03_MEE067_0390610_0391338 I mean what you do is you just build an X. M. L. file, and if you want it to get down to the utterances, you'd go to the leaves.796AMI_EN2001a_H03_MEE067_0391338_0391652 And then if you wanted the next level up797AMI_EN2001a_H03_MEE067_0391652_0392676 you'd go to the parents of those and like just go from like the leaves inwards towards the branch to build up things like um you know,798AMI_EN2001a_H03_MEE067_0392676_0393213 when you click on a segment, it's gonna have like words or whatever that are important.799AMI_EN2001a_H03_MEE067_0393998_0394605 As long as like the algorithms are designed um with it in mind, I don't think it's a very big problem.800AMI_EN2001a_H03_MEE067_0394919_0395920 Well like say you had um like say for a meeting, right, you've got like uh say a hierarchy that looks quite big, like this.801AMI_EN2001a_H03_MEE067_0396184_0396419 And like the utterances come off of here maybe.802AMI_EN2001a_H03_MEE067_0396604_0397214 Then when whatever your algorithm is doing, as long as when you're working with utterances, you go for all the leaves,803AMI_EN2001a_H03_MEE067_0397214_0397829 like then if you need something next up, so like a topic segment, you'd go to here.804AMI_EN2001a_H03_MEE067_0397829_0398181 But if you were looking at say this one, so only went like this.805AMI_EN2001a_H03_MEE067_0398518_0398948 Right, so you it's same, you'd start with the leaves, and you go oh, I want a topic segment.806AMI_EN2001a_H03_MEE067_0398948_0399048 So I go one layer up.807AMI_EN2001a_H03_MEE067_0399216_0399833 See, and then if you're working with just a topic segment in there, it's the only thing you have to worry about.808AMI_EN2001a_H03_MEE067_0399833_0400256 And like each time you want a higher level, you just need to go up the tree.809AMI_EN2001a_H03_MEE067_0400256_0400607 And as long as your algorithm respects that, then we can just810AMI_EN2001a_H03_MEE067_0400828_0401011 process any arbitrary X. M. L.811AMI_EN2001a_H03_MEE067_0401147_0401387 file with whatever hierarchical structure we want.812AMI_EN2001a_H03_MEE067_0401930_0402170 A meeting, say, and that would be a topic segment.813AMI_EN2001a_H03_MEE067_0402992_0403797 So I think as long as you build an algorithm that respects whatever structure's in the file, rather than imposing its own structure814AMI_EN2001a_H03_MEE067_0404796_0404913 Well no, it doesn't have to be.815AMI_EN2001a_H03_MEE067_0405120_0405381 But I mean it could be as many nodes as you want.816AMI_EN2001a_H03_MEE067_0405381_0405534 Like this one could be deeper maybe,817AMI_EN2001a_H03_MEE067_0405743_0405802 say.818AMI_EN2001a_H03_MEE067_0405802_0406582 So then you'd start with all your utterances here, and when you go up to get topic segments, you go to here here here here here here here.819AMI_EN2001a_H03_MEE067_0406800_0407099 That might be a bit confusing though 'cause you have things on different levels.820AMI_EN2001a_H03_MEE067_0410634_0410843 Well Wednesday.821AMI_EN2001a_H03_MEE067_0410843_0410884 Yeah.822AMI_EN2001a_H03_MEE067_0411256_0411282 Yeah.823AMI_EN2001a_H03_MEE067_0411485_0412181 So we'll see if we can get like a mini-browser just displays two things synched together of some kind.824AMI_EN2001a_H03_MEE067_0412181_0412299 Yeah.825AMI_EN2001a_H03_MEE067_0412299_0412314 Yeah.826AMI_EN2001a_H03_MEE067_0413368_0413430 It'd be useful.827AMI_EN2001a_H03_MEE067_0413430_0413599 I don't know who you see about that though.828AMI_EN2001a_H03_MEE067_0414992_0415103 I d have no idea.829AMI_EN2001a_H03_MEE067_0415396_0416113 I've probably got a reasonable amount because um everything on my DICE account can actually be deleted 'cause I store it all at home as well.830AMI_EN2001a_H03_MEE067_0417282_0417420 Is that guaranteed to stay, the831AMI_EN2001a_H03_MEE067_0418321_0418496 Maybe you should send a support form.832AMI_EN2001a_H03_MEE067_0418671_0418842 Just say we want some web space.833AMI_EN2001a_H03_MEE067_0419106_0419257 Listen to.834AMI_EN2001a_H03_MEE067_0419587_0419658 Yeah.835AMI_EN2001a_H03_MEE067_0419658_0420017 'Cause that'd be really useful is if we had a big directory.836AMI_EN2001a_H03_MEE067_0420017_0420237 Especially for transferring stuff.837AMI_EN2001a_H03_MEE067_0420672_0421047 Having said that, are we allowed to take a copy of the ICSI corpus?838AMI_EN2001a_H03_MEE067_0421047_0421230 Something we should probably ask before we do it.839AMI_EN2001a_H03_MEE067_0421701_0421753 Okay.840AMI_EN2001a_H03_MEE067_0423347_0423389 Okay.841AMI_EN2001a_H03_MEE067_0424527_0425014 No, me neither.842AMI_EN2001a_H03_MEE067_0441237_0441989 Might be funny to see what is summarised the whole corpus as anyway.843AMI_EN2001a_H03_MEE067_0441989_0442150 I think it'd be very useful.844AMI_EN2001a_H03_MEE067_0442150_0442160 But845AMI_EN2001a_H03_MEE067_0455372_0455482 We can just change the code.846AMI_EN2001a_H03_MEE067_0473718_0473789 Is that it?847AMI_EN2001a_H03_MEE067_0473789_0473855 That's quite good.848AMI_EN2001a_H03_MEE067_0478127_0478164 Yeah.849AMI_EN2001a_H03_MEE067_0481691_0482308 I could just use it with the frequency, I think, until the information density thing's finished.850AMI_EN2001a_H03_MEE067_0482308_0482418 That would be really useful.851AMI_EN2001a_H03_MEE067_0492163_0492803 If you're doing it in Java, could you um serialize the output as well as writing it to a file?852AMI_EN2001a_H03_MEE067_0492803_0493456 If you're doing it in Java, could you serialize the um dictionary, yeah, as well as writing it to a file?853AMI_EN2001a_H03_MEE067_0493625_0493752 It's really easy.854AMI_EN2001a_H03_MEE067_0494191_0494399 I don't see why it'd be any more massive than the file.855AMI_EN2001a_H03_MEE067_0494732_0494772 Yeah.856AMI_EN2001a_H03_MEE067_0495116_0495335 It just saves you parsing the um857AMI_EN2001a_H03_MEE067_0495546_0495671 file representation of it.858AMI_EN2001a_H03_MEE067_0495671_0495951 And now 'cause I would be using it in Java anyway.859AMI_EN2001a_H03_MEE067_0495951_0496174 So I'd just be building the data structure again.860AMI_EN2001a_H03_MEE067_0497185_0497672 Yeah, but it seems like a bit silly to be parsing it over and over again kinda thing.861AMI_EN2001a_H03_MEE067_0498725_0499449 I would've thought that um I think all the collections and things implement serializable already.862AMI_EN2001a_H03_MEE067_0499718_0499825 I think they might do.863AMI_EN2001a_H03_MEE067_0501302_0502109 Tonight I'll try and um I'll either work some more on uh the T. F. I. D. F. summarizer or do the audio thing.864AMI_EN2001a_H03_MEE067_0504214_0504258 Yeah.865AMI_EN2001a_H03_MEE067_0504724_0504888 Do we have to demonstrate something next week?866AMI_EN2001a_H03_MEE067_0506718_0506755 Yeah.867AMI_EN2001a_H03_MEE067_0509234_0509327 Yeah, I know.868AMI_EN2001a_H03_MEE067_0509829_0510797 I think it's 'cause we had to specify it ourselves that it's not as um like focus the specification of most um work we have to do.869AMI_EN2001a_H03_MEE067_0511362_0511409 Yeah.870AMI_EN2001a_H03_MEE067_0511546_0511950 Once we start doing it it will all become more or less obvious I think anyway.871AMI_EN2001a_H04_MEO069_0000334_0000388 'Kay.872AMI_EN2001a_H04_MEO069_0000560_0000601 Gosh.873AMI_EN2001a_H04_MEO069_0001048_0001088 'Kay.874AMI_EN2001a_H04_MEO069_0001677_0001805 Is there much more in it than he d875AMI_EN2001a_H04_MEO069_0001901_0002077 Is there much more in it than he said yesterday?876AMI_EN2001a_H04_MEO069_0002298_0002327 Mm.877AMI_EN2001a_H04_MEO069_0002508_0002702 Hmm.878AMI_EN2001a_H04_MEO069_0002702_0002744 Hmm?879AMI_EN2001a_H04_MEO069_0004683_0005624 Yeah, now I'd say if for the prototype if we just like wherever possible p chunk in the stuff that we have um pre-annotated and stuff880AMI_EN2001a_H04_MEO069_0005624_0005970 and for the stuff that we don't have pre-annotated write like a stupid baseline881AMI_EN2001a_H04_MEO069_0005970_0006441 then we should probably be able to basically that means we focus on on the interface first sort of882AMI_EN2001a_H04_MEO069_0006441_0007098 so that we we take the the ready-made parts and just see how we get them work together in the interface the way we want and and then we have a working prototype.883AMI_EN2001a_H04_MEO069_0007098_0007829 And then we can go back and replace pieces either by our own components or by more sophisticated compo po components of our own.884AMI_EN2001a_H04_MEO069_0007829_0007991 So it's probably feasible.885AMI_EN2001a_H04_MEO069_0007991_0008261 The thing is I'm away this weekend.886AMI_EN2001a_H04_MEO069_0008261_0008508 So that's for me887AMI_EN2001a_H04_MEO069_0009323_0009394 Oh yeah, um yeah.888AMI_EN2001a_H04_MEO069_0009394_0009419 No.889AMI_EN2001a_H04_MEO069_0009419_0010003 But also I might like the the similarity thing, like my just my matrix itself for my stuff890AMI_EN2001a_H04_MEO069_0010003_0010426 I c I I think I can do that fairly quickly because I have the algorithms.891AMI_EN2001a_H04_MEO069_0010426_0011043 Yeah, I think today's meeting is really the one where we where we sort of settle down the data structure and as soon as we have that,892AMI_EN2001a_H04_MEO069_0011043_0011289 uh probably like after today's meeting, we then actually need to893AMI_EN2001a_H04_MEO069_0011460_0012177 well go back first of all and look at NITE X. M. L. to see in how far that that which we want is compatible with that which NITE X. M. L. offers us.894AMI_EN2001a_H04_MEO069_0012177_0012665 And then just sort of everyone make sure everyone understand the interface.895AMI_EN2001a_H04_MEO069_0012911_0013629 So I think if today we decide on what data we wanna have now, and and later, maybe even today896AMI_EN2001a_H04_MEO069_0013629_0013961 we go and look at NITE X. M. L. or some of us look at NITE X. M. L. in a bit more detail897AMI_EN2001a_H04_MEO069_0013961_0014534 just trying to make some sense of that code and see how does the representation work in their system.898AMI_EN2001a_H04_MEO069_0014534_0015141 And then sort of with that knowledge we should be able to then say okay, that type of NITE X. M. L. data we wanna load into it,899AMI_EN2001a_H04_MEO069_0015141_0015737 and this is how everyone can access it, and then we should be able to go from there.900AMI_EN2001a_H04_MEO069_0015979_0016024 No.901AMI_EN2001a_H04_MEO069_0016157_0017488 I've looked looked at the documentation and n like seen enough to make me think that we want to use the NITE X. M. L. framework because um they have a good a event model that synchronizes sort of the data and and every display element.902AMI_EN2001a_H04_MEO069_0017488_0017761 So that takes a lot of work away from us.903AMI_EN2001a_H04_MEO069_0017761_0018162 Sort of that would be a reason for staying within their framework and using their general classes.904AMI_EN2001a_H04_MEO069_0018162_0018710 But beyond that I haven't looked at it at all, which is something we should really do.905AMI_EN2001a_H04_MEO069_0018710_0019626 Who actually like for this whole discussion I mean, who of us is doing stuff that is happening on-line and who of us is doing stuff that's happening off-line?906AMI_EN2001a_H04_MEO069_0019626_0019757 Like my data is coming c907AMI_EN2001a_H04_MEO069_0021058_0021074 Hmm?908AMI_EN2001a_H04_MEO069_0021316_0021363 Yeah.909AMI_EN2001a_H04_MEO069_0021842_0021886 Okay.910AMI_EN2001a_H04_MEO069_0022100_0022149 Okay.911AMI_EN2001a_H04_MEO069_0022897_0022930 'Kay.912AMI_EN2001a_H04_MEO069_0023174_0023651 So basically apart from the display module, the i the display itself913AMI_EN2001a_H04_MEO069_0023651_0024471 we don't have an extremely high degree of interaction between sort of our modules that create the stuff and and the interface,914AMI_EN2001a_H04_MEO069_0024471_0025049 so the interface is mainly while it's running just working on data that's just loaded from a file, I guess.915AMI_EN2001a_H04_MEO069_0025225_0025273 There isn't916AMI_EN2001a_H04_MEO069_0025788_0025836 Yeah, I know.917AMI_EN2001a_H04_MEO069_0025836_0026585 Th Yeah, the search is I guess the search is sort of a strange beast anyway because for the search we're leaving the NITE X. M. L. framework.918AMI_EN2001a_H04_MEO069_0026585_0026933 Um but that's still sort of that's good.919AMI_EN2001a_H04_MEO069_0026933_0027825 That means that at least like we don't have the type of situation where somebody has to do like a billion calculations on on data on-line.920AMI_EN2001a_H04_MEO069_0027825_0028518 'Cause that would make it a lot more like that would mean that our interface for the data would have to be a lot more careful about how it performs and and everything.921AMI_EN2001a_H04_MEO069_0028518_0028968 And nobody is modifying that data at at on-line time at all it seems.922AMI_EN2001a_H04_MEO069_0028968_0029312 Nobody's making any changes to the actual data on-line.923AMI_EN2001a_H04_MEO069_0029564_0029716 So that's actually making it a lot easier.924AMI_EN2001a_H04_MEO069_0029716_0030029 That basically means our browser really is a viewer mostly925AMI_EN2001a_H04_MEO069_0030029_0030614 which isn't doing much with the data except for sort of selecting a piece piece of it and and displaying it.926AMI_EN2001a_H04_MEO069_0031339_0031358 Hmm?927AMI_EN2001a_H04_MEO069_0031612_0031886 Well some parts relevant for the search, yes.928AMI_EN2001a_H04_MEO069_0031886_0031983 I'd say so.929AMI_EN2001a_H04_MEO069_0032583_0032608 Hmm?930AMI_EN2001a_H04_MEO069_0034338_0034715 Yeah, but nobody of us is doing much of searching from the data in the on-line stage.931AMI_EN2001a_H04_MEO069_0034882_0035121 And for all together, like the display itself,932AMI_EN2001a_H04_MEO069_0035121_0035759 I think we are easier if we if it's sitting on the X. M. L. than if it's sitting on the S. Q. L. stuff, because if it's sitting on the X. M. L.933AMI_EN2001a_H04_MEO069_0035759_0036968 we have the the NITE X. M. L. framework with all its functionality for synchronizing through all the different levels whenever there's a change, whenever something's moving forward and stuff.934AMI_EN2001a_H04_MEO069_0036968_0037277 And we can just more or less look at their code, like how their player moves forward935AMI_EN2001a_H04_MEO069_0037277_0037573 and how that moving forward is represented in different windows and stuff.936AMI_EN2001a_H04_MEO069_0037573_0038440 So I think in the actual browser itself I don't wanna sit on the S. Q. L. if we can sit on the X. M. L. because sitting on the X. M. L. we have all we have so much help.937AMI_EN2001a_H04_MEO069_0038440_0039285 And for y for like the p the calculations that we're doing apart from the search, it seems that everyone needs some special representations anyway.938AMI_EN2001a_H04_MEO069_0040012_0040165 You mean our results?939AMI_EN2001a_H04_MEO069_0040165_0040829 Yeah, in in the NITE X. M. L. X. M. L. format, so with their time stamps and stuff, so that it's easy to to tie together st things.940AMI_EN2001a_H04_MEO069_0041168_0041711 What I'm like what we have to think about is if we go with this multi-level idea,941AMI_EN2001a_H04_MEO069_0041711_0042569 like this idea that sort of if you start with a whole meeting series as one entity, as one thing that you display, as one whole sort of,942AMI_EN2001a_H04_MEO069_0042569_0043255 that then the individual chunks of the individual meetings, whereas and then you can click on a meeting943AMI_EN2001a_H04_MEO069_0043255_0043779 and then sort of the meeting is the whole thing and the chunks are the individual segments,944AMI_EN2001a_H04_MEO069_0043779_0044242 that means sort of we have multiple levels of of representation, which we probably945AMI_EN2001a_H04_MEO069_0044388_0044746 If we if we do it this way like we f we have to discuss that if we do it this way946AMI_EN2001a_H04_MEO069_0044746_0044987 then we should probably find some abstraction model947AMI_EN2001a_H04_MEO069_0044987_0046015 so that the interface in the sense like deals with it as if it's same so that the interface doesn't really have to worry whether it's a meeting in the whole meeting series or a segment within a meeting948AMI_EN2001a_H04_MEO069_0046015_0046186 you know what I mean?949AMI_EN2001a_H04_MEO069_0046186_0046500 And that's probably stuff that we have to sort of like process twice then.950AMI_EN2001a_H04_MEO069_0046500_0047294 Like for example that like the summary of a meeting within the whole meeting corpus or meeting series y is meeting series a good word for that?951AMI_EN2001a_H04_MEO069_0047294_0047426 I don't really know what how to call it.952AMI_EN2001a_H04_MEO069_0047426_0048009 You know what I mean, like not not the whole corpus, but every meeting that has to do with one topic.953AMI_EN2001a_H04_MEO069_0048009_0048470 Um so in in the meeting se series so that a summary for a meeting within the meeting series, are sort of954AMI_EN2001a_H04_MEO069_0048661_0048916 compiled off-line by a summary module.955AMI_EN2001a_H04_MEO069_0048916_0049376 And that is separate from a summary of a segment within a meeting.956AMI_EN2001a_H04_MEO069_0049376_0049466 'Cause I don't think we can957AMI_EN2001a_H04_MEO069_0052921_0053066 So are we doing that at all levels?958AMI_EN2001a_H04_MEO069_0053066_0053152 Are we um959AMI_EN2001a_H04_MEO069_0053867_0054209 And just have different like fine-grainedness levels sort of.960AMI_EN2001a_H04_MEO069_0054838_0054873 Mm.961AMI_EN2001a_H04_MEO069_0055179_0055233 'Kay.962AMI_EN2001a_H04_MEO069_0055233_0055966 So the only thing that yeah, so the only thing that would happen basically if I double-click let's say from the whole meeting series on a single meeting963AMI_EN2001a_H04_MEO069_0055966_0056164 is that the zoom level changes.964AMI_EN2001a_H04_MEO069_0056164_0056566 Like the th the start and the end position changes and the zoom level changes.965AMI_EN2001a_H04_MEO069_0056809_0057041 I I thought we couldn't do that.966AMI_EN2001a_H04_MEO069_0057041_0057400 Like I was under the impression that we couldn't do that because we couldn't load the data for all that.967AMI_EN2001a_H04_MEO069_0057604_0057723 But I don't know, I mean that968AMI_EN2001a_H04_MEO069_0059647_0059865 So I'm s not sure if I got it.969AMI_EN2001a_H04_MEO069_0059865_0059903 I was970AMI_EN2001a_H04_MEO069_0060512_0060622 Mm-hmm.971AMI_EN2001a_H04_MEO069_0060622_0060687 Mm-hmm.972AMI_EN2001a_H04_MEO069_0061386_0061412 Mm-hmm.973AMI_EN2001a_H04_MEO069_0061788_0061816 Mm-hmm.974AMI_EN2001a_H04_MEO069_0062077_0062113 Okay.975AMI_EN2001a_H04_MEO069_0062264_0062335 So976AMI_EN2001a_H04_MEO069_0062669_0062809 Okay.977AMI_EN2001a_H04_MEO069_0062809_0063102 I wa I was just worried about the total memory complexity of it.978AMI_EN2001a_H04_MEO069_0063102_0063796 But I I completely admit, I mean, I just sort of like th took that from some thing that Jonathan once said about not loading everything.979AMI_EN2001a_H04_MEO069_0063796_0063925 But maybe I was just wrong about it.980AMI_EN2001a_H04_MEO069_0063925_0064019 How many utterances w981AMI_EN2001a_H04_MEO069_0065185_0065436 Yeah, and I w yeah.982AMI_EN2001a_H04_MEO069_0065436_0065590 Yeah.983AMI_EN2001a_H04_MEO069_0065590_0065678 Yeah.984AMI_EN2001a_H04_MEO069_0065678_0065763 Yeah.985AMI_EN2001a_H04_MEO069_0065763_0066047 So what we have is we would have a word.986AMI_EN2001a_H04_MEO069_0066232_0066569 Like we would have words with some priority levels.987AMI_EN2001a_H04_MEO069_0066569_0067627 And they would basically be because even the selection would would the summaries automatically feed from just how prioritized an individual word or how indiv uh prioritized an individual utterance is?988AMI_EN2001a_H04_MEO069_0067627_0068143 Or i are the summaries sort of refined from it and made by a machine to make sentences and stuff?989AMI_EN2001a_H04_MEO069_0068143_0068648 Or are they just sort of taking out the words with the highest priority and then the words of the second highest priority?990AMI_EN2001a_H04_MEO069_0068831_0068989 And the u okay.991AMI_EN2001a_H04_MEO069_0069206_0069395 Are we doing it on th the whole thing on the utterance level?992AMI_EN2001a_H04_MEO069_0069395_0069670 Or are we doing it on word level, like the information density calculation?993AMI_EN2001a_H04_MEO069_0070390_0070736 We I think we have start and end times for words actually, but it's yeah994AMI_EN2001a_H04_MEO069_0070736_0070973 but it m it might s but it might sound crazy in the player.995AMI_EN2001a_H04_MEO069_0070973_0071472 We should really maybe we can do that together at some point today that we check out how the player works.996AMI_EN2001a_H04_MEO069_0071472_0071820 But there's maybe some merit in altogether doing it on an utterance level in the end.997AMI_EN2001a_H04_MEO069_0072033_0072054 So998AMI_EN2001a_H04_MEO069_0073667_0073708 Yeah.999AMI_EN2001a_H04_MEO069_0074675_0075085 Well but also about the displays, I mean the displays in the in the text body1000AMI_EN2001a_H04_MEO069_0075085_0075553 in the in the latest draft that we had sort of we came up with the idea that it isn't displaying utterance for utterance,1001AMI_EN2001a_H04_MEO069_0075553_0076370 but it's also displaying uh a summarised version in you know, like below the below the graph, the part.1002AMI_EN2001a_H04_MEO069_0076370_0076471 Maybe1003AMI_EN2001a_H04_MEO069_0076649_0076694 Yeah, r1004AMI_EN2001a_H04_MEO069_0076858_0076876 Hmm?1005AMI_EN2001a_H04_MEO069_0076996_0077436 Oh yeah, f it's just like there there's like audio skimming and there's displayed skimming.1006AMI_EN2001a_H04_MEO069_0077706_0077882 Yeah.1007AMI_EN2001a_H04_MEO069_0077882_0078504 Ma maybe there's some merit of going altogether for utterance level and not even bother to calculate I mean if you have to do it internally1008AMI_EN2001a_H04_MEO069_0078504_0078549 then you can do it.1009AMI_EN2001a_H04_MEO069_0078549_0079271 But maybe like not even store the importance levels for individual words and just sort of rank utterances as a whole.1010AMI_EN2001a_H04_MEO069_0079568_0079586 Hmm?1011AMI_EN2001a_H04_MEO069_0079724_0079775 Yeah.1012AMI_EN2001a_H04_MEO069_0079923_0080330 'Cause it it might be better skimming and less memory required at the same time.1013AMI_EN2001a_H04_MEO069_0080330_0080723 And I mean if you if you know how to do it for individual words, then you can just in the worst case,1014AMI_EN2001a_H04_MEO069_0080723_0081057 if you can't find anything else, just sort of make the mean of the words over the utterance.1015AMI_EN2001a_H04_MEO069_0081057_0081113 You know what I mean?1016AMI_EN2001a_H04_MEO069_0081313_0081467 W it's it's1017AMI_EN2001a_H04_MEO069_0081686_0082248 Well what's the smallest chunk at the moment you're thinking of of assigning an importance measure to, is it a word or is it an utterance?1018AMI_EN2001a_H04_MEO069_0082510_0082994 So we're thinking of like maybe just storing it on a per utterance level.1019AMI_EN2001a_H04_MEO069_0083220_0083571 Because it's it's less stuff to store probably for Dave in the in the audio playing.1020AMI_EN2001a_H04_MEO069_0083571_0083927 And for in the display it's probably better if you have whole utterances than I don't know1021AMI_EN2001a_H04_MEO069_0083927_0084374 like what it's like if you just take single words out of utterances.1022AMI_EN2001a_H04_MEO069_0084374_0084912 That probably doesn't make any sense at all, whereas if you just uh show important utterances but the utterance as a whole it makes more sense.1023AMI_EN2001a_H04_MEO069_0084912_0085308 So it doesn't actually make a difference for your algorithm, 'cause it just means that if you're working on a word level1024AMI_EN2001a_H04_MEO069_0085308_0085600 then we just mean it over the utterance.1025AMI_EN2001a_H04_MEO069_0086519_0086578 They are on1026AMI_EN2001a_H04_MEO069_0086937_0087075 Oh so that's good anyway then, yeah.1027AMI_EN2001a_H04_MEO069_0087075_0087319 Because that makes it a lot easier than to t put it on utterance level.1028AMI_EN2001a_H04_MEO069_0087958_0087994 Oh yeah.1029AMI_EN2001a_H04_MEO069_0087994_0088294 No but I mean like how how Jasmine does it internally I don't know,1030AMI_EN2001a_H04_MEO069_0088294_0088580 but it's probably, yeah, you probably have to work on word levels for importance.1031AMI_EN2001a_H04_MEO069_0088580_0089083 But there should be ways of easily going from a word level to an utterance level.1032AMI_EN2001a_H04_MEO069_0089083_0089117 Okay.1033AMI_EN2001a_H04_MEO069_0090106_0090151 Yeah, prob1034AMI_EN2001a_H04_MEO069_0090658_0090688 Hmm.1035AMI_EN2001a_H04_MEO069_0093127_0093349 Well we do a pre-filtering of sort of the whole thing,1036AMI_EN2001a_H04_MEO069_0093349_0093712 sort of like but that, like the problem with that is it's easy to do in the text level.1037AMI_EN2001a_H04_MEO069_0093712_0094446 But that would mean it would still play the uh in your audio, unless we sort of also store what pieces we cut out for the audio.1038AMI_EN2001a_H04_MEO069_0094879_0094993 Yeah.1039AMI_EN2001a_H04_MEO069_0094993_0095433 I think before we can like answer that specific question how we c deal with that1040AMI_EN2001a_H04_MEO069_0095433_0095803 it's probably good for us to look at what the audio player is capable of doing.1041AMI_EN2001a_H04_MEO069_0096033_0096081 Yes.1042AMI_EN2001a_H04_MEO069_0096447_0096554 So what do you mean by buffering?1043AMI_EN2001a_H04_MEO069_0096554_0096730 Like you think directly feeding1044AMI_EN2001a_H04_MEO069_0097129_0097681 But yeah, but not but not stored on the hard disk and then loaded in, but loaded in directly from memory.1045AMI_EN2001a_H04_MEO069_0098261_0098896 But it's probably a stream if it exists in Java, it would be probably some binary stream going in of some type.1046AMI_EN2001a_H04_MEO069_0098896_0098985 Okay, yeah.1047AMI_EN2001a_H04_MEO069_0099586_0099696 Okay.1048AMI_EN2001a_H04_MEO069_0099696_0100150 Okay, so I mean so that means that there's probably, even if you go on an per utterance level,1049AMI_EN2001a_H04_MEO069_0100150_0100741 there's still some merit on within utterances cutting out stuff which clearly isn't relevant at all, and that maybe also for the audio we'd have to do.1050AMI_EN2001a_H04_MEO069_0100741_0101536 So let's say we play the whole au phrase, but then in addition to that, we have some information that says minus that part of something.1051AMI_EN2001a_H04_MEO069_0101536_0101650 That's okay, that we can do.1052AMI_EN2001a_H04_MEO069_0103214_0103633 Yeah, maybe even I mean that's sort of that depends on how how advanced we get.1053AMI_EN2001a_H04_MEO069_0103633_0104019 If maybe if we realise that there's massive differences in in gain or in something,1054AMI_EN2001a_H04_MEO069_0104019_0104613 you can probably just make some simple simple normalization, but that really depends on how much time we have and and how much is necessary.1055AMI_EN2001a_H04_MEO069_0104613_0104958 Yeah, if like I d I don't know anything about audio and I have never seen the player.1056AMI_EN2001a_H04_MEO069_0104958_0105645 So if you find that the player accepts some n input from memory, and if it's easy to do, then I guess that's that's fairly doable.1057AMI_EN2001a_H04_MEO069_0105645_0105848 So but that means in the general structure we're actually quite lucky1058AMI_EN2001a_H04_MEO069_0105848_0106916 so we we have we load into memory for the whole series of meetings just the utterances and rankings for the utterances and some information probably that says,1059AMI_EN2001a_H04_MEO069_0106916_0107170 well, the I guess that goes with the utterance, who's speaking.1060AMI_EN2001a_H04_MEO069_0107353_0107576 Because then we can also do the display about who's speaking.1061AMI_EN2001a_H04_MEO069_0108020_0108063 Yeah.1062AMI_EN2001a_H04_MEO069_0108365_0109244 But I'm I'm still confused 'cause I thought like that's just what Jonathan said we do c that we can't do, like load a massive document of that size.1063AMI_EN2001a_H04_MEO069_0109244_0109319 On the other hand1064AMI_EN2001a_H04_MEO069_0110014_0110547 The other hand, I mean it shouldn't be like should be like fifty mega-byte in RAM or something, it shouldn't be massive, should it?1065AMI_EN2001a_H04_MEO069_0110547_0110764 Actually fifty hundred megabyte is quite big in RAM.1066AMI_EN2001a_H04_MEO069_0111599_0111834 Just thinking, what's the simp so1067AMI_EN2001a_H04_MEO069_0112011_0112650 We do get an error message with the project if we load everything into the project with all the data they load.1068AMI_EN2001a_H04_MEO069_0112650_0112818 So we know that doesn't work.1069AMI_EN2001a_H04_MEO069_0112818_0113168 So our hope is essentially that we load less into it.1070AMI_EN2001a_H04_MEO069_0113168_0113483 What's this lazy loading thing, somebody explain lazy loading to me.1071AMI_EN2001a_H04_MEO069_0114200_0114333 Ah, okay.1072AMI_EN2001a_H04_MEO069_0114333_0114581 So that is that only by type of file.1073AMI_EN2001a_H04_MEO069_0114581_0115056 Like if if if the same thing is in different files, would it then maybe like, you know1074AMI_EN2001a_H04_MEO069_0115056_0115633 if if utterances are split over three or ten or w hundred different files,1075AMI_EN2001a_H04_MEO069_0115633_0116009 is then a chance maybe that it doesn't try to load them all into memory at the same time, but just1076AMI_EN2001a_H04_MEO069_0116340_0116507 So why does it fail then in the first place?1077AMI_EN2001a_H04_MEO069_0116507_0116733 Then it shouldn't ever fail, because then it should never1078AMI_EN2001a_H04_MEO069_0116951_0117710 Yeah, but yeah, but um it uh it it failed right when you load it, right, the NITE X. M. L. kit, so that's interesting.1079AMI_EN2001a_H04_MEO069_0119247_0119290 Hmm.1080AMI_EN2001a_H04_MEO069_0120444_0120583 Let's check that out.1081AMI_EN2001a_H04_MEO069_0120583_0120926 Um I'll p I'll probably ask Jonathan about it.1082AMI_EN2001a_H04_MEO069_0120926_0121370 So alternatively, if we realise we can't do the whole thing in one go,1083AMI_EN2001a_H04_MEO069_0121370_0121992 we can probably just process some sort of meta-data, you know what I mean, like sort of sort of for the whole series1084AMI_EN2001a_H04_MEO069_0122159_0122574 chunks representing the individual meetings or some1085AMI_EN2001a_H04_MEO069_0122764_0123754 Like something that represents the whole series in in a v in a structure very similar to the structure in which we represent individual um meetings1086AMI_EN2001a_H04_MEO069_0123754_0124222 but with data sort of always combined from the whole series.1087AMI_EN2001a_H04_MEO069_0124222_0125153 so instead of having an single utterance that we display, it would probably be like that would be representing a whole um topic, a segment in a meeting.1088AMI_EN2001a_H04_MEO069_0125330_0125531 And sort of so that wi using the same data st1089AMI_EN2001a_H04_MEO069_0126264_0126333 Well, in a sense1090AMI_EN2001a_H04_MEO069_0127218_0127804 Uh I'm I'm thinking of in a sense of like creating a virtual a virtual meeting out of the whole meeting series, sort of.1091AMI_EN2001a_H04_MEO069_0128172_0128897 Yeah, sort of like off-line create a virtual meeting, which which basically treats the meeting series as if it was a meeting,1092AMI_EN2001a_H04_MEO069_0128897_0130026 and treats the individual meetings within the series as if they were segments, and treats the individual segments within meetings as if they were um utterances.1093AMI_EN2001a_H04_MEO069_0130026_0130325 You know, so we just sort of we shift it one level up.1094AMI_EN2001a_H04_MEO069_0130325_0130882 And in that way we could probably use the same algorithm and just like make vir like one or two ifs that say okay1095AMI_EN2001a_H04_MEO069_0130882_0131346 if you are on a whole document uh a whole series level and that was a double-click,1096AMI_EN2001a_H04_MEO069_0131346_0131976 then don't just go into that um segment, but load a new file or something like it, but in general use the same algorithm.1097AMI_EN2001a_H04_MEO069_0131976_0132222 That would be an alternative if we can't actually load the whole thing and1098AMI_EN2001a_H04_MEO069_0132480_0133236 'Cause also like even if we maybe this whole like maybe I'm worrying too much about the whole series in one thing display1099AMI_EN2001a_H04_MEO069_0133236_0133618 because actually I mean probably users wouldn't view that one too often.1100AMI_EN2001a_H04_MEO069_0136453_0136599 Yeah, but I'm I'm still worried.1101AMI_EN2001a_H04_MEO069_0136599_0137220 Like for example for the display, if you actually if you want a display uh like for the whole series1102AMI_EN2001a_H04_MEO069_0137220_0137825 the information density levels based on and and the f and the only granularity you have is individual utterances1103AMI_EN2001a_H04_MEO069_0137825_0138410 that means you have to through every single utterance in a series of seventy hours of meetings.1104AMI_EN2001a_H04_MEO069_0138935_0139102 Yeah.1105AMI_EN2001a_H04_MEO069_0139102_0139615 Yeah, and if you make that structurally very similar to the the le like one level down1106AMI_EN2001a_H04_MEO069_0139615_0139834 like the way how we uh store individual utterances and stuff1107AMI_EN2001a_H04_MEO069_0139834_0140209 then maybe we can more or less use the same code and just make a few ifs and stuff.1108AMI_EN2001a_H04_MEO069_0140544_0140993 Yeah, so so but still so in in general we're having we're having utterances1109AMI_EN2001a_H04_MEO069_0141169_0141314 and they have a score.1110AMI_EN2001a_H04_MEO069_0141541_0141928 And that's as much as we really need.1111AMI_EN2001a_H04_MEO069_0141928_0142419 And of cou and they also have a time a time information of course.1112AMI_EN2001a_H04_MEO069_0142419_0142492 Hmm?1113AMI_EN2001a_H04_MEO069_0142492_0142685 And a and a s and a speaker information, yeah.1114AMI_EN2001a_H04_MEO069_0143079_0143300 Yeah, so an information which topic they're in, yeah.1115AMI_EN2001a_H04_MEO069_0143300_0143643 And and probably separate to that an information about the different topics like that1116AMI_EN2001a_H04_MEO069_0143834_0143866 Yeah.1117AMI_EN2001a_H04_MEO069_0144018_0144100 So1118AMI_EN2001a_H04_MEO069_0144239_0144900 so the skimming can work on that because the skimming just sort of sorts the utterances and puts as many in as it needs1119AMI_EN2001a_H04_MEO069_0145423_0145515 Yeah.1120AMI_EN2001a_H04_MEO069_0145515_0146152 Yeah, it'll it'll play them in some order in which they were set because otherwise it's gonna be more entertaining.1121AMI_EN2001a_H04_MEO069_0146152_0146200 Um1122AMI_EN2001a_H04_MEO069_0146361_0147114 but that that's enough data for the skimming and the the searching, so what the searching does is the searching leaves the whole framework,1123AMI_EN2001a_H04_MEO069_0147114_0148063 goes to the S. Q. L. database and gets like basically in the end gets just a time marker for where that is, like that utterance that we are concerned with.1124AMI_EN2001a_H04_MEO069_0148063_0148671 And then we have to find I'm sure there's some way in in NITE X. M. L. to just say set position to that time mark.1125AMI_EN2001a_H04_MEO069_0148671_0149139 And then it shifts the whole frame and it alerts every single element of the display and the display updates.1126AMI_EN2001a_H04_MEO069_0149594_0149729 Yeah, yeah.1127AMI_EN2001a_H04_MEO069_0149729_0150018 That we can ju yeah, but so so if if somethi so yeah.1128AMI_EN2001a_H04_MEO069_0150018_0150263 So if in that tree display somebody clicks on something1129AMI_EN2001a_H04_MEO069_0150519_0150946 Yeah, and then you sort of feed the time stamp to and the NITE X. M. L. central manager,1130AMI_EN2001a_H04_MEO069_0150946_0151518 and that central manager alerts everything that's there, like alerts the skim like the the audio display1131AMI_EN2001a_H04_MEO069_0151518_0151624 alerts the text display1132AMI_EN2001a_H04_MEO069_0151624_0152326 alerts the visual display and says we have a new time frame and then they all sort of do their update routines with respect to the current level of zoom.1133AMI_EN2001a_H04_MEO069_0152326_0152938 So how much do they display, and starting position at where the or maybe the mid-position of it,1134AMI_EN2001a_H04_MEO069_0152938_0153515 I don't know, like w if start where the thing was found or if that thing wa was found it's in the middle of the part that we display1135AMI_EN2001a_H04_MEO069_0153515_0153569 that I don't know.1136AMI_EN2001a_H04_MEO069_0153569_0153775 But that we can decide about, but a general sort of1137AMI_EN2001a_H04_MEO069_0154010_0154584 It's the same thing if like whether you play and it moves forward or whether you jump to a position through search,1138AMI_EN2001a_H04_MEO069_0154584_0154957 it's essentially for all the window handling, it's the same event.1139AMI_EN2001a_H04_MEO069_0154957_0155654 It's only that the event gets triggered by the search routine which sort of push that into NITE X. M. L. and says please go there now.1140AMI_EN2001a_H04_MEO069_0157742_0157937 Why do we have to do it in memory?1141AMI_EN2001a_H04_MEO069_0157937_0158012 But that stuff's so1142AMI_EN2001a_H04_MEO069_0158269_0158502 I mean like the information is coming from off-line.1143AMI_EN2001a_H04_MEO069_0158831_0159339 So we probably we don't even have to change the utterance document,1144AMI_EN2001a_H04_MEO069_0159339_0159765 right, because the whole way, like the whole beauty of the NITE X. M. L. is that it ties together lots of different files.1145AMI_EN2001a_H04_MEO069_0159765_0160373 So we can just create an additional X. M. L. file which for every utterance like the utterances have I. D. s I presume, some references.1146AMI_EN2001a_H04_MEO069_0160373_0160722 So we just we tie uh p just a very short X. M. L. file1147AMI_EN2001a_H04_MEO069_0160722_0161372 which it's the only information it has that has whatever a number for for the um weight,1148AMI_EN2001a_H04_MEO069_0161372_0161949 for the information density, and we just tie that to the existing utterances and tie them to the existing speaker changes.1149AMI_EN2001a_H04_MEO069_0164681_0165696 Well otherwise we probably have to go over it and like add some integer that we just increment from top to bottom sort of to every utterance as an as an I. D. some type.1150AMI_EN2001a_H04_MEO069_0165696_0166371 Or un or try to understand how NITE X. M. L. I. D. s work and maybe there's some special route we have to follow when we use these I. D. s.1151AMI_EN2001a_H04_MEO069_0166371_0166524 It's alm hmm?1152AMI_EN2001a_H04_MEO069_0167049_0167350 Yeah, the the girl said the utterances themselves are not numbered at the moment.1153AMI_EN2001a_H04_MEO069_0167587_0167727 Okay.1154AMI_EN2001a_H04_MEO069_0167727_0167889 Okay.1155AMI_EN2001a_H04_MEO069_0167889_0167987 Okay.1156AMI_EN2001a_H04_MEO069_0167987_0168067 Yeah.1157AMI_EN2001a_H04_MEO069_0168067_0168285 So I guess that would be solvable if not.1158AMI_EN2001a_H04_MEO069_0168704_0168740 Mm-hmm.1159AMI_EN2001a_H04_MEO069_0169383_0169414 Sorry?1160AMI_EN2001a_H04_MEO069_0170085_0170111 Okay.1161AMI_EN2001a_H04_MEO069_0170298_0170324 Okay.1162AMI_EN2001a_H04_MEO069_0170672_0170825 Is that a board marker pen actually?1163AMI_EN2001a_H04_MEO069_0171140_0171237 Oh.1164AMI_EN2001a_H04_MEO069_0171237_0171757 That's just so like to make a list of all this stuff, or we probably can somebody can do it on paper.1165AMI_EN2001a_H04_MEO069_0171941_0172087 All these fancy pens.1166AMI_EN2001a_H04_MEO069_0172087_0173107 So what so the stuff we have we have utterances and speakers and weights for utterances.1167AMI_EN2001a_H04_MEO069_0173107_0173726 So for for every utterance sort of like the utterance has a speaker and a weight which is coming from outside.1168AMI_EN2001a_H04_MEO069_0173726_0173942 Or we just tie it to it.1169AMI_EN2001a_H04_MEO069_0173942_0174440 And there is segments, which hmm?1170AMI_EN2001a_H04_MEO069_0174736_0174869 Oh, so sorry um.1171AMI_EN2001a_H04_MEO069_0174869_0175029 Uh topic s topic segments I meant.1172AMI_EN2001a_H04_MEO069_0175029_0175174 Like they are they are a super-unit.1173AMI_EN2001a_H04_MEO069_0175174_0175528 So so the utterances are tied to topic segments.1174AMI_EN2001a_H04_MEO069_0176171_0176880 And if the time stamps are on a word level, then we b somehow have to extract time stamps for utterances where they start.1175AMI_EN2001a_H04_MEO069_0177437_0177578 W what segments now?1176AMI_EN2001a_H04_MEO069_0178271_0178391 Okay.1177AMI_EN2001a_H04_MEO069_0178391_0178891 Is the uh is that the same as utterances that is that the same as utterances that1178AMI_EN2001a_H04_MEO069_0179825_0179967 Mm-hmm.1179AMI_EN2001a_H04_MEO069_0179967_0179996 Mm-hmm.1180AMI_EN2001a_H04_MEO069_0181252_0181565 What so that's Oh.1181AMI_EN2001a_H04_MEO069_0181565_0181951 But that's one o one segment or is that two segments then?1182AMI_EN2001a_H04_MEO069_0182368_0182402 Yeah.1183AMI_EN2001a_H04_MEO069_0182700_0182824 Okay.1184AMI_EN2001a_H04_MEO069_0182824_0182845 Okay.1185AMI_EN2001a_H04_MEO069_0182845_0183574 So but but generally utterances is that which we just called uh sorry, segments is that which we just called utterances now.1186AMI_EN2001a_H04_MEO069_0183574_0184006 Like it's it's the sa it's sort of like one person's contribution at a time sort of thingy dingy.1187AMI_EN2001a_H04_MEO069_0184188_0184692 Okay, so yeah, so we have those, and and then we have some f field somewhere else which has1188AMI_EN2001a_H04_MEO069_0184869_0184979 topics.1189AMI_EN2001a_H04_MEO069_0185119_0185540 Yeah, and and a topic's basically they are just on the I. D., probably with a start time or something1190AMI_EN2001a_H04_MEO069_0185540_0185939 and and the utterances referenced to those topics I guess.1191AMI_EN2001a_H04_MEO069_0185939_0186403 So the topics don't contain any redundant thing of like showing the whole topic again1192AMI_EN2001a_H04_MEO069_0186403_0186827 but they just sort of say a number and where they start and where they finish.1193AMI_EN2001a_H04_MEO069_0186827_0187036 And the utterances then say which topic they belong to.1194AMI_EN2001a_H04_MEO069_0188457_0188479 Yeah.1195AMI_EN2001a_H04_MEO069_0188479_0188509 No.1196AMI_EN2001a_H04_MEO069_0188509_0188986 But I was thinking of the topic segmentation now and and f for that there would only be one,1197AMI_EN2001a_H04_MEO069_0188986_0189200 right, because it's sort of like it's just a time window.1198AMI_EN2001a_H04_MEO069_0189388_0189418 Yeah.1199AMI_EN2001a_H04_MEO069_0189602_0190423 So if this lazy loading works, then this should definitely fit into I mean not memory then because it wouldn't all be in memory at the same time.1200AMI_EN2001a_H04_MEO069_0190423_0191454 So if we just have those sort of that information like a long list of all the utterances slash segments and like short or smaller lists which give weight to them.

Showing the first 1,200 of 108503 lines. Download the file for the rest.