text.md

Text Modules

Modules pre-trained to embed words, phrases, and sentences as many-dimensional vectors.

Click on a module to view its documentation, or reference the URL from the TensorFlow Hub library like so:

m = hub.Module("https://tfhub.dev/...")

Universal Sentence Encoder

Encoder of greater-than-word length text trained on a variety of data.

• universal-sentence-encoder
• universal-sentence-encoder-large
• universal-sentence-encoder-lite (*Text preprocessing required)

ELMo

Deep Contextualized Word Representations trained on the 1 Billion Word Benchmark.

• elmo

NNLM embedding trained on Google News

Embedding from a neural network language model trained on the Google News dataset.

	50 dimensions	128 dimensions
Chinese	nnlm-zh-dim50 nnlm-zh-dim50-with-normalization	nnlm-zh-dim128 nnlm-zh-dim128-with-normalization
English	nnlm-en-dim50 nnlm-en-dim50-with-normalization	nnlm-en-dim128 nnlm-en-dim128-with-normalization
German	nnlm-de-dim50 nnlm-de-dim50-with-normalization	nnlm-de-dim128 nnlm-de-dim128-with-normalization
Indonesian	nnlm-id-dim50 nnlm-id-dim50-with-normalization	nnlm-id-dim128 nnlm-id-dim128-with-normalization
Japanese	nnlm-ja-dim50 nnlm-ja-dim50-with-normalization	nnlm-ja-dim128 nnlm-ja-dim128-with-normalization
Korean	nnlm-ko-dim50 nnlm-ko-dim50-with-normalization	nnlm-ko-dim128 nnlm-ko-dim128-with-normalization
Spanish	nnlm-es-dim50 nnlm-es-dim50-with-normalization	nnlm-es-dim128 nnlm-es-dim128-with-normalization

Word2vec trained on Wikipedia

Embedding trained by word2vec on Wikipedia.

English

250 dimensions	500 dimensions
Wiki-words-250 Wiki-words-250-with-normalization	Wiki-words-500 Wiki-words-500-with-normalization