UNAL-NLP: Cross-Lingual Phrase Sense Disambiguation with Syntactic Dependency Trees

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 743–747, Dublin, Ireland, August 23-24, 2014. UNAL-NLP: Cross-Lingual Phrase Sense Disambiguation with Syntactic Dependency Trees Emilio Silva-Schlenker Departamento de Lingüística Universidad Nacional de Colombia Departamento de Ingeniería de Sistemas Universidad de los Andes, Bogotá D.C., Colombia esilvas@unal.edu.co Sergio Jimenez and Julia Baquero Universidad Nacional de Colombia, Bogotá D.C., Colombia sgjimenezv@unal.edu.co jmbauqerov@unal.edu.co Abstract In this paper we describe our participa- tion in the SemEval 2014, Task 5, con- sisting of the construction of a translation assistance system that translates L1 frag- ments, written in L2 context, to their cor- rect L2 translation. Our approach con- sists of a bilingual parallel corpus, a sys- tem of syntactic features extraction and a statistical memory-based classification al- gorithm. Our system ranked 4th and 6th among the 10 participating systems that used the English-Spanish data set. 1 Introduction An L2 writing assistant is a tool intended for lan- guage learners who need to improve their writing skills. This tool lets them write a text in L2, but fall back to their native L1 whenever they are not sure about a certain word or expression. In these cases, the assistant automatically translates this text for them (van Gompel et al., 2014). Although at first glance this may be seen as a classification problem, it might be better ful- filled by a cross-lingual word sense disambigua- tion (WSD) approach, which takes context into account by means of contextual features used in a machine learning setting. The main differences between this and previous approaches to cross- lingual WSD are the bilingual nature of the input sentences (see section 2.3) and the annotation of target phrases, rather than single words. The remainder of this article is organized as fol- lows. Section 2 describes the proposed method. A description of the system we submitted, the ob- tained results and an error analysis are discussed This work is licensed under a Creative Commons At- tribution 4.0 International Licence. Page numbers and pro- ceedings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/4.0/ in section 3. In section 4 we present a brief dis- cussion about the results. Finally, in section 5 we make some concluding remarks. 2 Method Description The core of the proposed system uses techniques from memory-based classification to find the most appropriate translation of a target phrase in a given context. It receives an input as in (1) and yields an output as in (2). (1) No creo que ella is coming. (2) No creo que ella venga. It does so on the basis of a syntactic selec- tion of context features, a large bilingual parallel corpus and a classifier built using the Tilburg Memory-Based Learner, TiMBL (Daelemans et al., 2010). The proposed system consists of several stages. First, a large bilingual corpus is aligned at word and phrase level. Next, an index is built by each phrase in the L1 side of the corpus to retrieve ef- ficiently the occurrences of a particular L1 phrase …

Content truncated to 500 words