SERGIOJIMENEZ at SemEval-2016 Task 1: Effectively Combining Paraphrase Database, String Matching, WordNet, and Word Embedding for Semantic Textual Similarity

Proceedings of SemEval-2016, pages 749–757, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics SERGIOJIMENEZ at SemEval-2016 Task 1: Effectively Combining Paraphrase Database, String Matching, WordNet and Word Embedding for Semantic Textual Similarity Sergio Jimenez Bogotá D.C., Colombia sergio.jimenez.vargas@gmail.com Abstract In this paper, a system for semantic tex- tual similarity, which participated in Task- 1 in SemEval 2016 (monolingual and cross- lingual sub-tasks) is described. The system contains a preprocessing step that simplifies text using PPDB 2.0 and detects negations. Also, six lexical similarity functions were constructed using string matching, word em- bedding and synonyms-antonyms relations in WordNet. These lexical similarity functions are projected to sentence level using a new method called Polarized Soft Cardinality that supports negative similarities between words to model opposites. We also introduce a novel L2-norm “cardinality” for vector space repre- sentations. The system extracts a set of 660 features from each pair of text snippets using the proposed cardinality measures. From this set, a subset of 12 features was selected in a supervised manner. These features are com- bined by SVR and, alternatively, by using the arithmetic mean to produce similarity predic- tions. Our team ranked second in the cross- lingual sub-task and got close to the best offi- cial results in the monolingual sub-task. 1 Introduction Semantic Textual Similarity (STS) is a fundamental task in the field of natural language processing that has been addressed in SemEval competitions unin- terruptedly since 2012 (Agirre et al., 2012; Agirre et al., 2013; Agirre et al., 2014; Agirre et al., 2015). The task is to compare two text fragments and pro- duce a similarity score that is assessed according to human judgment. This year (Agirre et al., 2016), a new cross-lingual sub-task in English and Spanish is proposed in addition to the traditional monolingual English task. In SemEval 2015, the most popular ap- proach among the best systems was the use of words alignments between sentences combining resources such as WordNet (Miller, George A., 1995), neural word embedding (Mikolov, Tomas et al., 2013) and the Paraphrase Database (Pavlick et al., 2015). Figure 1: STS system architecture Polarized soft cardinality Text snippet A PPDB Paraphrase simplification Tokenizing & stop-words removal Negation detection Text snippet B bags of words Lexical similarity word pairwise similarities vector mapper vectors L2-norm cardinality Feature combination Feature selection SVR STS data 2012-2015 average combination Preprocessing Feature extraction Model learning GloVe word2vec Processing pipeline Resources 2.1 WordNet 2.2 2.2 3 4 5 6 7 8 8 3.2 run1 run2 run3 749 This paper describes our system submission to STS 2016 that uses a cardinality-based approach (instead of word alignments) for combining the resources mentioned above. Several teams have used soft cardinality successfully in previous STS competitions from 2012 to 2014 (Jimenez et al., 2012; Jimenez et al., 2013a; Jimenez et al., 2013b; Jimenez et al., 2014; Lynum et al., 2014). For the proposed system, we extended the model of soft car- dinality to allow the use of negative values in the lex- ical …

Content truncated to 500 words