Back to the Feature, in Entailment Detection and Similarity Measurement for Portuguese

Abstract

This paper describes a system to identify entailment and quantify semantic similarity among pairs of Portuguese sentences. The system relies on a corpus to build a supervised model, and employs the same features regardless of the task. Our experiments cover two types of features, contextualized embeddings and lexical features, which we eval- uate separately and in combination. The model is derived from a voting strategy on an ensemble of distinct regressors, on similarity measure- ment, or calibrated classifiers, on entailment detection. Applying such system to other languages mainly depends on the availability of cor- pora, since all features are either multilingual or language independent. We obtain competitive results on a recent Portuguese corpus, where our best result is obtained by joining embeddings with lexical features.