def get_word2vec(lang: str = "en"):
# Download.
urls = {
"en": "https://s3.amazonaws.com/dl4j-distribution/GoogleNews-vectors-negative300.bin.gz",
"ja": "http://public.shiroyagi.s3.amazonaws.com/latest-ja-word2vec-gensim-model.zip"
}
path = download.cached_download(urls[lang])
path = Path(path)
filename = "word2vec.gensim.model"
print("Loading model...")
if lang == "ja":
dirpath = Path(download.get_cache_directory(str(Path("word2vec"))))
download.cached_unzip(path, dirpath / lang)
model_path = dirpath / lang / filename
model = gensim.models.Word2Vec.load(str(model_path))
if lang == "en":
dirpath = Path(download.get_cache_directory(str(Path("word2vec") / "en")))
model_path = dirpath / filename
download.cached_decompress_gzip(path, model_path)
model = gensim.models.KeyedVectors.load_word2vec_format(str(model_path), binary=True)
return model
def get_fasttext(lang: str = "en"):
# Download.
urls = {
"en": "https://dl.fbaipublicfiles.com/fasttext/vectors-wiki/wiki.simple.zip",
"ja": "https://dl.fbaipublicfiles.com/fasttext/vectors-wiki/wiki.ja.zip",
"fr": "https://dl.fbaipublicfiles.com/fasttext/vectors-wiki/wiki.fr.zip",
}
path = download.cached_download(urls[lang])
path = Path(path)
dirpath = path.parent / 'fasttext' / lang
download.cached_unzip(path, dirpath)
print("Loading model...")
filename = Path(urls[lang]).stem + '.bin'
model = load_model(str(dirpath / filename))
return model
def test_cache_unzip(self):
# TODO: Not tested if the zipfile is actually unziped.
saveto = Path(self.temp_dir) / "saveto"
download.cached_unzip(Path(self.zippath), saveto=saveto)