View Code? Open in Web Editor
NEW
This project forked from dhwajraj /deep-siamese-text-similarity
Tensorflow based implementation of deep siamese LSTM network to capture sentence similarity using word embeddings
License: MIT License
Python 45.92%
Jupyter Notebook 44.88%
Shell 9.20%
deep-siamese-text-similarity's People
Watchers
deep-siamese-text-similarity's Issues
STS zip packed do not include:
STS2016 - cross-lingual English-Spanish (trial, test) datasets
STS2015 - Spanish (train, test) datasets
STS2014 - Spanish (train, test) datasets
Prepare updated package containing mentioned datasets. Place it in publicly available place.
Test various configurations:
different RNN stack parameters (hidden units, layers, dropout)
(optional) different word2vec models
different train / test datasets
Spearman's rho coefficient seems to be not converging
The training procedure should be parametrizable with two different word2vec models (side1_word2vec and side2_word2vec). The model should be able to work with different languages.