RUFINO at SemEval-2017 Task 2: Cross-lingual lexical similarity by extending PMI and word embeddings systems with a Swadesh’s-like list

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 239–244, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics RUFINO at SemEval-2017 Task 2: Cross-lingual lexical similarity by extending PMI and word embeddings systems with a Swadesh’s-like list Sergio Jimenez Instituto Caro y Cuervo Bogotá, Colombia sergio.jimenez @caroycuervo .gov.co George Dueñas Instituto Caro y Cuervo Bogotá, Colombia george.duenas @caroycuervo .gov.co Lorena Gaitan Universidad Cooperativa de Colombia Bogotá, Colombia lorena.gaitanb @campusucc.edu.co Jorge Segura Universidad Nacional de Colombia Bogotá, Colombia jaseguran @unal.edu.co Abstract The RUFINO team proposed a non- supervised, conceptually-simple and low- cost approach for addressing the Multi- lingual and Cross-lingual Semantic Word Similarity challenge at SemEval 2017. The proposed systems were cross-lingual extensions of popular monolingual lex- ical similarity approaches such as PMI and word2vec. The extensions were pos- sible by means of a small parallel list of concepts similar to the Swadesh’s list, which we obtained in a semi-automatic way. In spite of its simplicity, our ap- proach showed to be effective obtaining statistically-significant and consistent re- sults in all datasets proposed for the task. Besides, we provide some research direc- tions for improving this novel and afford- able approach. 1 Introduction Pairwise semantic lexical similarity is a core com- ponent in NLP systems that tackle fundamental NLP tasks such as word sense disambiguation (Camacho-Collados et al., 2015), semantic textual similarity (Agirre et al., 2017) and many others. Since more than two decades, the problem has been addressed mainly for the English language, but only recently, other languages have been con- sidered. The task 2 in SemEval 2017 (Camacho- Collados et al., 2017) proposes a public challenge for this task in 5 languages (English, Spanish, Ital- ian, German and Farsi) and an additional cross- lingual challenge in their 10 possible combina- tions. This paper describes the participating sys- tems of the RUFINO team in these challenges. Lexical-similarity systems receive two words as input and return a numerical score that reflects the Lexical-similarity system in English Lexical-similarity system in Spanish s1 s2 s3 … sn Cosine similarity wEN= “volleyball” SIM (wEN, wES) wES= “pelota” language-transverse concepts volleyball pelota eye tree blood ojo árbol sangre s1 s2 s3 … sn cEN,n cES,n s3 ... . . . .. . . wES wEN Figure 1: Architecture of the cross-lingual system similarity or relatedness between them. Cross- lingual systems extends the idea to words in dif- ferent languages. The evaluation of such systems consist in measuring the correlation of the scores obtained by several word pairs against the consen- sus of human judgments (gold standard). The main fundamental resources used by lexical similarity systems are monolingual corpora, par- allel corpora and knowledge-based resources such as WordNet (Miller, 1995) and Babelnet (Navigli and Ponzetto, 2012). Among them, monolingual corpora are the cheapest and most available re- source in the majority of the languages. Aiming to propose a lexical similarity system with easy 239 replicability, beyond the 5 languages of the chal- lenge, the RUFINO team proposed a system based …

Content truncated to 500 words