Evaluation and classification of syntax information usage in determining short text semantic similarity

This paper outlines and categorizes ways of using syntax information in a number of algorithms for determining short text semantic similarity. Algorithm performance was evaluated using the results of a paraphrase detection test on the Microsoft Research Paraphrase Corpus. Among the described algorithms and approaches to using syntax information we identify those best suited for application in languages with limited electronic linguistic tools and, with that goal in mind, we propose a new algorithm classification.