semeru/code-text-javascript

1---2license: mit3Programminglanguage: "JavaScript"4version: "N/A"5Date: "Codesearchnet(Jun 2020 - paper release date)"6Contaminated: "Very Likely"7Size: "Standar Tokenizer (TreeSitter)"8---9 10### Dataset is imported from CodeXGLUE and pre-processed using their script.11 12# Where to find in Semeru:13The dataset can be found at /nfs/semeru/semeru_datasets/code_xglue/code-to-text/javascript in Semeru14 15 16# CodeXGLUE -- Code-To-Text17 18## Task Definition19 20The task is to generate natural language comments for a code, and evaluted by [smoothed bleu-4](https://www.aclweb.org/anthology/C04-1072.pdf) score.21 22## Dataset23 24The dataset we use comes from [CodeSearchNet](https://arxiv.org/pdf/1909.09436.pdf) and we filter the dataset as the following:25 26- Remove examples that codes cannot be parsed into an abstract syntax tree.27- Remove examples that #tokens of documents is < 3 or >25628- Remove examples that documents contain special tokens (e.g. <img ...> or https:...)29- Remove examples that documents are not English.30 31 32### Data Format33 34After preprocessing dataset, you can obtain three .jsonl files, i.e. train.jsonl, valid.jsonl, test.jsonl35 36For each file, each line in the uncompressed file represents one function.  One row is illustrated below.37 38  - **repo:** the owner/repo39 40  - **path:** the full path to the original file41 42  - **func_name:** the function or method name43 44  - **original_string:** the raw string before tokenization or parsing45 46  - **language:** the programming language47 48  - **code/function:** the part of the `original_string` that is code49 50  - **code_tokens/function_tokens:** tokenized version of `code`51 52  - **docstring:** the top-level comment or docstring, if it exists in the original string53 54  - **docstring_tokens:** tokenized version of `docstring`55 56### Data Statistic57 58| Programming Language | Training |  Dev   |  Test  |59| :------------------- | :------: | :----: | :----: |60| JavaScript           |  58,025  | 3,885  | 3,291  |61 62 63## Reference64<pre><code>@article{husain2019codesearchnet,65  title={Codesearchnet challenge: Evaluating the state of semantic code search},66  author={Husain, Hamel and Wu, Ho-Hsiang and Gazit, Tiferet and Allamanis, Miltiadis and Brockschmidt, Marc},67  journal={arXiv preprint arXiv:1909.09436},68  year={2019}69}</code></pre>70