TangibleAI/mathtext-fastapi
1
1import re2 3from collections.abc import Mapping4from logging import getLogger5import datetime as dt6from dateutil.parser import isoparse7 8from fuzzywuzzy import fuzz9from fuzzywuzzy import process10from mathtext_fastapi.intent_classification import predict_message_intent11from mathtext_fastapi.logging import prepare_message_data_for_logging12from mathtext.sentiment import sentiment13from mathtext.text2int import text2int, TOKENS2INT_ERROR_INT14 15log = getLogger(__name__)16 17PAYLOAD_VALUE_TYPES = {18 'author_id': str,19 'author_type': str,20 'contact_uuid': str,21 'message_body': str,22 'message_direction': str,23 'message_id': str,24 'message_inserted_at': str,25 'message_updated_at': str,26 }27 28 29def build_nlu_response_object(nlu_type, data, confidence):30 """ Turns nlu results into an object to send back to Turn.io31 Inputs32 - nlu_type: str - the type of nlu run (integer or sentiment-analysis)33 - data: str/int - the student message34 - confidence: - the nlu confidence score (sentiment) or '' (integer)35 36 >>> build_nlu_response_object('integer', 8, 0)37 {'type': 'integer', 'data': 8, 'confidence': 0}38 39 >>> build_nlu_response_object('sentiment', 'POSITIVE', 0.99)40 {'type': 'sentiment', 'data': 'POSITIVE', 'confidence': 0.99}41 """42 return {43 'type': nlu_type,44 'data': data,45 'confidence': confidence46 }47 48 49# def test_for_float_or_int(message_data, message_text):50# nlu_response = {}51# if type(message_text) == int or type(message_text) == float:52# nlu_response = build_nlu_response_object('integer', message_text, '')53# prepare_message_data_for_logging(message_data, nlu_response)54# return nlu_response55 56 57def test_for_number_sequence(message_text_arr, message_data, message_text):58 """ Determines if the student's message is a sequence of numbers59 60 >>> test_for_number_sequence(['1','2','3'], {"author_id": "57787919091", "author_type": "OWNER", "contact_uuid": "df78gsdf78df", "message_body": "I am tired", "message_direction": "inbound", "message_id": "dfgha789789ag9ga", "message_inserted_at": "2023-01-10T02:37:28.487319Z", "message_updated_at": "2023-01-10T02:37:28.487319Z"}, '1, 2, 3')61 {'type': 'integer', 'data': '1,2,3', 'confidence': 0}62 63 >>> test_for_number_sequence(['a','b','c'], {"author_id": "57787919091", "author_type": "OWNER", "contact_uuid": "df78gsdf78df", "message_body": "I am tired", "message_direction": "inbound", "message_id": "dfgha789789ag9ga", "message_inserted_at": "2023-01-10T02:37:28.487319Z", "message_updated_at": "2023-01-10T02:37:28.487319Z"}, 'a, b, c')64 {}65 """66 nlu_response = {}67 if all(ele.isdigit() for ele in message_text_arr):68 nlu_response = build_nlu_response_object(69 'integer',70 ','.join(message_text_arr),71 072 )73 prepare_message_data_for_logging(message_data, nlu_response)74 return nlu_response75 76 77def run_text2int_on_each_list_item(message_text_arr):78 """ Attempts to convert each list item to an integer79 80 Input81 - message_text_arr: list - a set of text extracted from the student message82 83 Output84 - student_response_arr: list - a set of integers (32202 for error code)85 86 >>> run_text2int_on_each_list_item(['1','2','3'])87 [1, 2, 3]88 """89 student_response_arr = []90 for student_response in message_text_arr:91 int_api_resp = text2int(student_response.lower())92 student_response_arr.append(int_api_resp)93 return student_response_arr94 95 96def run_sentiment_analysis(message_text):97 """ Evaluates the sentiment of a student message98 99 >>> run_sentiment_analysis("I am tired")100 [{'label': 'NEGATIVE', 'score': 0.9997807145118713}]101 102 >>> run_sentiment_analysis("I am full of joy")103 [{'label': 'POSITIVE', 'score': 0.999882698059082}]104 """105 # TODO: Add intent labelling here106 # TODO: Add logic to determine whether intent labeling or sentiment analysis is more appropriate (probably default to intent labeling)107 return sentiment(message_text)108 109 110def run_intent_classification(message_text):111 """ Process a student's message using basic fuzzy text comparison112 113 >>> run_intent_classification("exit")114 {'type': 'intent', 'data': 'exit', 'confidence': 1.0}115 >>> run_intent_classification("exi") 116 {'type': 'intent', 'data': 'exit', 'confidence': 0.86}117 >>> run_intent_classification("eas")118 {'type': 'intent', 'data': '', 'confidence': 0}119 >>> run_intent_classification("hard")120 {'type': 'intent', 'data': '', 'confidence': 0}121 >>> run_intent_classification("hardier") 122 {'type': 'intent', 'data': 'harder', 'confidence': 0.92}123 """124 label = ''125 ratio = 0126 nlu_response = {'type': 'intent', 'data': label, 'confidence': ratio}127 keywords = [128 'easier',129 'exit',130 'harder',131 'hint',132 'next',133 'stop',134 'tired',135 'tomorrow',136 'finished',137 'help',138 'easier',139 'easy',140 'support',141 'skip',142 'menu'143 ]144 145 try:146 tokens = re.findall(r"[-a-zA-Z'_]+", message_text.lower())147 except AttributeError:148 tokens = ''149 150 for keyword in keywords:151 try:152 tok, score = process.extractOne(keyword, tokens, scorer=fuzz.ratio)153 except:154 score = 0155 156 if score > 80:157 nlu_response['data'] = keyword158 nlu_response['confidence'] = score159 160 return nlu_response161 162 163def payload_is_valid(payload_object):164 """165 >>> payload_is_valid({'author_id': '+5555555', 'author_type': 'OWNER', 'contact_uuid': '3246-43ad-faf7qw-zsdhg-dgGdg', 'message_body': 'thirty one', 'message_direction': 'inbound', 'message_id': 'SDFGGwafada-DFASHA4aDGA', 'message_inserted_at': '2022-07-05T04:00:34.03352Z', 'message_updated_at': '2023-04-06T10:08:23.745072Z'})166 True167 168 >>> payload_is_valid({"author_id": "@event.message._vnd.v1.chat.owner", "author_type": "@event.message._vnd.v1.author.type", "contact_uuid": "@event.message._vnd.v1.chat.contact_uuid", "message_body": "@event.message.text.body", "message_direction": "@event.message._vnd.v1.direction", "message_id": "@event.message.id", "message_inserted_at": "@event.message._vnd.v1.chat.inserted_at", "message_updated_at": "@event.message._vnd.v1.chat.updated_at"})169 False170 """171 try:172 isinstance(173 isoparse(payload_object.get('message_inserted_at','')),174 dt.datetime175 )176 isinstance(177 isoparse(payload_object.get('message_updated_at','')),178 dt.datetime179 )180 except ValueError:181 return False182 return (183 isinstance(payload_object, Mapping) and184 isinstance(payload_object.get('author_id'), str) and185 isinstance(payload_object.get('author_type'), str) and186 isinstance(payload_object.get('contact_uuid'), str) and187 isinstance(payload_object.get('message_body'), str) and188 isinstance(payload_object.get('message_direction'), str) and189 isinstance(payload_object.get('message_id'), str) and190 isinstance(payload_object.get('message_inserted_at'), str) and191 isinstance(payload_object.get('message_updated_at'), str) 192 )193 194 195def log_payload_errors(payload_object):196 errors = []197 try:198 assert isinstance(payload_object, Mapping)199 except Exception as e:200 log.error(f'Invalid HTTP request payload object: {e}')201 errors.append(e)202 for k, typ in PAYLOAD_VALUE_TYPES.items():203 try:204 assert isinstance(payload_object.get(k), typ)205 except Exception as e:206 log.error(f'Invalid HTTP request payload object: {e}')207 errors.append(e)208 try:209 assert isinstance(210 dt.datetime.fromisoformat(payload_object.get('message_inserted_at')),211 dt.datetime212 )213 except Exception as e:214 log.error(f'Invalid HTTP request payload object: {e}')215 errors.append(e)216 try: 217 isinstance(218 dt.datetime.fromisoformat(payload_object.get('message_updated_at')),219 dt.datetime220 )221 except Exception as e:222 log.error(f'Invalid HTTP request payload object: {e}')223 errors.append(e)224 return errors225 226 227def evaluate_message_with_nlu(message_data):228 """ Process a student's message using NLU functions and send the result229 230 >>> evaluate_message_with_nlu({"author_id": "57787919091", "author_type": "OWNER", "contact_uuid": "df78gsdf78df", "message_body": "8", "message_direction": "inbound", "message_id": "dfgha789789ag9ga", "message_inserted_at": "2023-01-10T02:37:28.487319Z", "message_updated_at": "2023-01-10T02:37:28.487319Z"})231 {'type': 'integer', 'data': 8, 'confidence': 0}232 233 >>> evaluate_message_with_nlu({"author_id": "57787919091", "author_type": "OWNER", "contact_uuid": "df78gsdf78df", "message_body": "I am tired", "message_direction": "inbound", "message_id": "dfgha789789ag9ga", "message_inserted_at": "2023-01-10T02:37:28.487319Z", "message_updated_at": "2023-01-10T02:37:28.487319Z"})234 {'type': 'sentiment', 'data': 'NEGATIVE', 'confidence': 0.9997807145118713}235 """236 # Keeps system working with two different inputs - full and filtered @event object237 # Call validate payload238 log.info(f'Starting evaluate message: {message_data}')239 240 if not payload_is_valid(message_data):241 log_payload_errors(message_data)242 return {'type': 'error', 'data': TOKENS2INT_ERROR_INT, 'confidence': 0}243 244 try:245 message_text = str(message_data.get('message_body', ''))246 except:247 log.error(f'Invalid request payload: {message_data}')248 # use python logging system to do this//249 return {'type': 'error', 'data': TOKENS2INT_ERROR_INT, 'confidence': 0}250 251 # Run intent classification only for keywords252 intent_api_response = run_intent_classification(message_text)253 if intent_api_response['data']:254 prepare_message_data_for_logging(message_data, intent_api_response)255 return intent_api_response256 257 number_api_resp = text2int(message_text.lower())258 259 if number_api_resp == TOKENS2INT_ERROR_INT:260 # Run intent classification with logistic regression model261 predicted_label = predict_message_intent(message_text)262 if predicted_label['confidence'] > 0.01:263 nlu_response = predicted_label264 else:265 # Run sentiment analysis266 sentiment_api_resp = sentiment(message_text)267 nlu_response = build_nlu_response_object(268 'sentiment',269 sentiment_api_resp[0]['label'],270 sentiment_api_resp[0]['score']271 )272 else:273 nlu_response = build_nlu_response_object(274 'integer',275 number_api_resp,276 0277 )278 279 prepare_message_data_for_logging(message_data, nlu_response)280 return nlu_response281