/sif-java

Implementation of ICLR 2017 "sentence embedding by Smooth Inverse Frequency weighting scheme" in Java.

Primary LanguageJava

Description

Implementation of SIF sentence embedding in Java.

  • Use pre-computed tfidf weights & word2vec to compute weighted sentence embedding.
  • Use PCA to compute principle components for all pairs of sentences and remove it.

For usage see test cases.

Data

Please download word2vec to data. I used glove.6B.50d.txt from Glove.