doc2vec-api

The repository contains some corpus(Korean), python scripts for training and inferring test document vectors using doc2vec.

Demo Site

• korean word2vec/doc2vec demo site

Raw Corpus

• Korean Wikipedia / space tokenizer (467MB)
• Korean Wikipedia / mecab pos tokenizer / tag info (910MB)
• Korean Wikipedia / mecab pos tokenizer / no tag info (535MB)

PreTrained Doc2vec Model

• Korean Wikipedia / mecab pos tokenizer / no tag info / 30 vectors(dmpv)

• Korean Wikipedia / mecab pos tokenizer / no tag info / 100 vectors(dmpv)

• Korean Wikipedia / mecab pos tokenizer / no tag info / 300 vectors(dmpv)

• Korean Wikipedia / mecab pos tokenizer / no tag info / 1000 vectors(dmpv)

• Korean Wikipedia + financial news / mecab pos tokenizer / no tag info / 30 vectors(dmpv)

• Korean Wikipedia + financial news / mecab pos tokenizer / no tag info / 100 vectors(dmpv)

Korean word2vec-api / doc2vec-api

Simple web service providing a word embedding API. The methods are based on Gensim Word2Vec / Doc2Vec implementation. Models are passed as parameters and must be in the Word2Vec / Doc2Vec text or binary format. This web2vec-api script is forked from this word2vec-api github and get minor update to support Korean word2vec models.

• Install Dependencies

pip2 install -r requirements.txt

• Launching the service

python word2vec-api --model path/to/the/model [--host host --port 1234]
ex) python /home/word2vec-api.py --model /home/model/all_terms_50vectors --path /word2vec --host 0.0.0.0 --port 4000


python doc2vec-api --model path/to/the/model [--host host --port 1234]
ex) python /home/doc2vec-api.py --model /home/model/all_terms_50vectors --path /doc2vec --host 0.0.0.0 --port 4000

• Example calls

curl http://127.0.0.1:5000/word2vec/most_similar?positive=무증