Skip to content

Implementation of a Siamese neural network in Keras for predicting sentence entailment.

License

Notifications You must be signed in to change notification settings

Lancasterg/siamese-lstm-sick-dataset

Repository files navigation

LSTM Siamese neural network for predicting sentence entailment

The Sentences Involving Compositional Knowledge (SICK) dataset consists of 9,840 pairs of sentences. Each sentence pair is labelled as either contradiction, neutral or entailment. This repo uses a deep, Siamese, bidirectional, Long Short-Term Memory (LSTM) network to predict sentence entailment using Word2Vec embeddings. The data set is split into 4,934 training pairs and 4,906 test pairs.

Usage

Run:

  • python controller.py

Author & Credit

This repo is an adaptation of the brilliant Siamese neural network implmentation by Aman Srivastava.

Releases

No releases published

Packages

No packages published

Languages