Assessing Word Similarity Metrics for Traceability Link Recovery - Evaluation Dataset

This dataset includes all data that was used for the evaluation of my bachelor's thesis:

Assessing Word Similarity Metrics for Traceability Link Recovery

The following files correspond to the following data sets from the evaluation:

cc-en-300.tar.gz corresponds to fastText's cc.en.300.bin embedding

crawl-300d-2M-subword.tar.gz corresponds to fastText's crawl-300d-2M-subword.bin embedding

wiki-news-300d-1M-subword.tar.gz corresponds to fastText's wiki-news-300d-1M-subword.bin embedding

wordnet.tar.gz corresponds to the WordNet 3.1 semantic network

sewordsim.tar.gz corresponds to SEWordSimDB's vector similarity database

glove_cc_840B_300d.tar.gz corresponds to GloVe's CC vector embedding

glove_wikigiga_300d.tar.gz corresponds to GloVe's 300 dimensional WIGI vector embedding

glove_wikigiga_200d.tar.gz corresponds to GloVe's 200 dimensional WIGI vector embedding

glove_wikigiga_100d.tar.gz corresponds to GloVe's 100 dimensional WIGI vector embedding

glove_wikigiga_50d.tar.gz corresponds to GloVe's 50 dimensional WIGI vector embedding

glove_twitter_200d.tar.gz corresponds to GloVe's 200 dimensional TWTR vector embedding

glove_twitter_100d.tar.gz corresponds to GloVe's 100 dimensional TWTR vector embedding

glove_twitter_50d.tar.gz corresponds to GloVe's 50 dimensional TWTR vector embedding

glove_twitter_25d.tar.gz corresponds to GloVe's 25 dimensional TWTR vector embedding

eval_results.tar.gz contains the detailed evaluation results for each configuration of all measures

The licenses of all data sets are included in their respective files.

Some of these data sets are .sql files. To use these files to reproduce the evaluation, they need to be imported into a sqlite3 database. The version of ArDoCo used for the evaluation is only able to work with sqlite3 databases and not with sql files.