Max-Cosine Matching Based Neural Models for Recognizing Textual Entailment
This work addresses a fundamental task in natural language processing for applications like text mining, but it is incremental as it builds on existing neural approaches with specific enhancements.
The paper tackled the Recognizing Textual Entailment (RTE) problem by proposing a neural model that matches words between hypothesis and premise using max-cosine similarity, enhanced with three techniques, and achieved improved predictive accuracy on the SNLI dataset, outperforming several state-of-the-art methods.
Recognizing textual entailment is a fundamental task in a variety of text mining or natural language processing applications. This paper proposes a simple neural model for RTE problem. It first matches each word in the hypothesis with its most-similar word in the premise, producing an augmented representation of the hypothesis conditioned on the premise as a sequence of word pairs. The LSTM model is then used to model this augmented sequence, and the final output from the LSTM is fed into a softmax layer to make the prediction. Besides the base model, in order to enhance its performance, we also proposed three techniques: the integration of multiple word-embedding library, bi-way integration, and ensemble based on model averaging. Experimental results on the SNLI dataset have shown that the three techniques are effective in boosting the predicative accuracy and that our method outperforms several state-of-the-state ones.