UQeResearch: Semantic Textual Similarity Quantification

Abstract

This paper presents an approach for estimating the Semantic Textual Similarity of full English sentences as specified in Shared Task 2 of SemEval-2015. The semantic similarity of sentence pairs is quantified from three perspectives - structural, syntactical, and semantic. The numerical representations of the derived similarity measures are then applied to train a regression ensemble. Although none of these three sets of measures is able to represent the semantic similarity of two sentences individually, our experimental results show that the combination of these features can precisely assess the semantic similarity of the sentences. In the English subtask our system's best result ranked 35 among 73 system runs with 0.7189 average Pearson correlation over five test sets. This was 0.08 correlation points less than the best submitted run.