Using Linguistic Data for English and Spanish Verb-Noun Combination Identification

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 857–867, Osaka, Japan, December 11-17 2016. Using Linguistic Data for English and Spanish Verb-Noun Combination Identification Uxoa I˜nurrieta, Arantza D´ıaz de Ilarraza, Gorka Labaka, Kepa Sarasola IXA NLP group, University of the Basque Country usoa.inurrieta|a.diazdeilarraza|gorka.labaka|kepa.sarasola@ehu.eus Itziar Aduriz Department of Linguistics, University of Barcelona itziar.aduriz@ub.edu John Carroll Department of Informatics, University of Sussex j.a.carroll@sussex.ac.uk Abstract We present a linguistic analysis of a set of English and Spanish verb+noun combinations (VNCs), and a method to use this information to improve VNC identification. Firstly, a sample of frequent VNCs are analysed in-depth and tagged along lexico-semantic and morphosyntactic dimensions, obtaining satisfactory inter-annotator agreement scores. Then, a VNC identification experiment is undertaken, where the analysed linguistic data is combined with chunking information and syntactic dependencies. A comparison between the results of the experiment and the results obtained by a basic detection method shows that VNC identification can be greatly improved by using linguistic information, as a large number of additional occurrences are detected with high precision. 1 Introduction Multiword Expressions (MWEs) are recurrent combinations of two or more words expressing a single unit of meaning, this meaning not always derivable directly from the meanings of the component words (Sag et al., 2002). Therefore, Natural Language Processing (NLP) tasks that need to be sensitive to lexical meaning should treat MWEs as single units. However, this is a challenging problem since many MWEs can have multimple morphosyntactic variants, which makes them difficult to recognise or gener- ate. Examples (1)-(3) below contain take steps; correct translation of this MWE into another language, for instance, requires it to be recognised as a single unit1. (1) The Government will take all the necessary steps to prepare. (2) They set out five important steps the Minister needs to take. (3) What were the steps that should have been taken? Although the most straightforward method for recognising MWEs is to attempt to match word se- quences against entries in a lexicon, this method does not work for combinations that can have multiple variants. This is often the situation for verb+noun combinations (VNCs), since this kind of MWE is usually morphosyntactically flexible. In the case of Machine Translation (MT), there are two challenges that need to be addressed concerning VNCs: (1) the detection of a given combination in the source language, and (2) its translation into the target language. If the first part fails, the words that constitute the MWE will be translated separately, This work is licensed under a Creative Commons Attribution 4.0 International Licence. Licence details: http:// creativecommons.org/licenses/by/4.0/ 1Google Translate English-French and English-Spanish https://translate.google.co.uk apparently detects take steps as an MWE in (1) but not in (2) or (3). 857 which will usually result in an incorrect translation. Then, for the second part, it is vital to have the necessary information to know what translation should be given to each VNC. A further problem arises here, since the morphosyntax of this kind of MWE varies a …

Content truncated to 500 words