Max-Cosine Matching Based Neural Models for Recognizing Textual Entailment

Abstract

Abstract. 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.

Enhancing Brazilian Portuguese Textual Entailment Recognition with a Hybrid Approach

“…Additionally, in agreement with the previous resources, the author uses antonymy, hypernymy, hyponymy and synonymy contained in WordNet to compose the attributes used in a Support Vector Machine (SVM) classifier. Xie and Hu (2017) presents an approach based on max-cosine matching for natural language inference in short sentences. In this approach, the first step involves word similarity evaluation and the next step is to represent this word pairs in order to apply a LSTM Artificial Neural Network architecture to identify the sentences similarity.…”