Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
This is an implementation of the LexVec word embedding model (similar to word2vec and GloVe) that achieves state of the art results in multiple NLP tasks
| Date | Stars |
|---|---|
| 2026-07-31 | 807 |
| 2026-08-03 | 807 |
| 2026-08-06 | 807 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# LexVec This is an implementation of the **LexVec word embedding model** (similar to word2vec and GloVe) that achieves state of the art results in multiple NLP tasks, as described in [these papers](#refs). ## Pre-trained Vectors ### Subword LexVec [(paper)](http://www.aclweb.org/anthology/W18-1209) * [Common Crawl](http://web-language-models.s3-website-us-east-1.amazonaws.com/wmt16/deduped/en-new.xz) - 58B tokens, cased - 2,000,000 words - 300 dimensions - [Word Vectors (2.1GB)](https://www.dropbox.com/s/mrxn933chn5u37z/lexvec.commoncrawl.ngramsubwords.300d.W.pos.vectors.gz?dl=1) - [Binary model (8.6GB)](https://www.dropbox.com/s/buix0deqlks4312/lexvec.commoncrawl.ngramsubwords.300d.W.pos.bin.gz?dl=1) - use this to [compute vectors for out-of-vocabulary (OOV) words](#oov) ### LexVec ([paper 1](http://anthology.aclweb.org/P16-2068), [paper 2](https://arxiv.org/pdf/1606.01283)) * [Common Crawl](http://web-language-models.s3-website-us-east-1.amazonaws.com/wmt16/deduped/en-new.xz) - 58B tokens, lowercased - 2,000,000 words - 300 dimensions - [Word Vectors (2.2GB)](https://www.dropbox.com/s/flh1fjynqvdsj4p/lexvec.commoncrawl.300d.W.pos.vectors.gz?dl=1) - [Word + Context Vectors (2.3GB)](https://www.dropbox.com/s/zkiajh6fj0hm0m7/lexvec.commoncrawl.300d.W%2BC.pos.vectors.gz?dl=1) * English Wikipedia 2015 + [NewsCrawl](http://www.statmt.org/wmt14/translation-task.html) - 7B tokens - 368,999 words - 300 dimensions - [Word Vectors (398MB)](https://www.dropbox.com/s/kguufyc2xcdi8yk/lexvec.enwiki%2Bnewscrawl.300d.W.pos.vectors.gz?dl=1) - [Word + Context Vectors (426MB)](https://www.dropbox.com/s/u320t9bw6tzlwma/lexvec.enwiki%2Bnewscrawl.300d.W%2BC.pos.vectors.gz?dl=1) ## Evaluation: Subword LexVec ### External memory, huge corpus | Model | GSem | GSyn | MSR | RW | SimLex | SCWS | WS-Sim | WS-Rel | MEN | MTurk | | ----- | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | | LexVec | 72.6% | **73.8%** | **73.2%** | **.539** | **.477** | **.687** | .809 | .696 | **.814** | **.717** | | fastText | **75.0%** | 72.1% | 71.8% | .522 | .424 | .673 | **.810** | **.724** | .805 | **.717** | * Both models use vectors with 300 dimensions. * Both models use character n-grams of length 3-6 as subwords. * All tasks are evaluated using cased words (``"Toronto" != "toronto"``). * GSem, GSyn, and MSR analogies were solved using [3CosMul](http://www.aclweb.org/anthology/W14-1618). * Both models were trained using [this release of Common Crawl](http://web-language-models.s3-website-us-east-1.amazonaws.com/wmt16/deduped/en-new.xz) which contains **58B tokens**, restricting the vocabulary to the 2 million most frequent cased words. * Subword LexVec was trained using the following command: ``` $ OUTPUT=output scripts/em_lexvec.sh -corpus common_crawl_cased.txt -negative 3 -dim 300 -subsample 1e-5 -minfreq 0 -window 2 -minn 3 -maxn 6 ``` * fastText was trained using the following command: ``` $ ./fasttext skipgram -input common_crawl_cased.txt -minCount 0 -t 1e-5 -dim 300 -lr 0.025 -minn 3 -maxn 6 ``` ## Evaluation: LexVec ### In-memory, large corpus | Model | GSem | GSyn | MSR | RW | SimLex | SCWS | WS-Sim | WS-Rel | MEN | MTurk | | ----- | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | | LexVec, Word | **81.1%** | **68.7%** | **63.7%** | **.489** | **.384** | **.652** | .727 | .619 | .759 | .655 | | LexVec, Word + Context | 79.3% | 62.6% | 56.4% | .476 | .362 | .629 | .734 | **.663** | **.772** | .649 | | word2vec Skip-gram | 78.5% | 66.1% | 56.0% | .471 | .347 | .649 | **.774** | .647 | .759 | **.687** | * All three models were trained using the same English Wikipedia 2015 + NewsCrawl corpus. * GSem, GSyn, and MSR analogies were solved using [3CosMul](http://www.aclweb.org/anthology/W14-1618). * LexVec was trained using the default parameters, expanded here for comparison: ``` $ OUTPUT=output sc
Excerpt of 9,737 characters
Read on GitHub
33
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:5b6aab7820dfa8f7, desc:embedding model