Towards the Automatic Merging of Lexical Resources: Automatic Mapping

Proceedings of Recent Advances in Natural Language Processing, pages 296–301, Hissar, Bulgaria, 12-14 September 2011. Towards the Automatic Merging of Lexical Resources: Automatic Mapping Muntsa Padró Núria Bel Silvia Necsulescu IULA Universitat Pompeu Fabra Barcelona, Spain IULA Universitat Pompeu Fabra Barcelona, Spain IULA Universitat Pompeu Fabra Barcelona, Spain muntsa.padro@upf.edu nuria.bel@upf.edu silvia.necsulescu@upf.edu Abstract Lexical Resources are a critical component for Natural Language Processing applications. However, the high cost of comparing and merg- ing different resources has been a bottleneck to have richer resources with a broad range of po- tential uses for a significant number of lan- guages. With the objective of reducing cost by eliminating human intervention, we present a new method for automating the merging of re- sources, with special emphasis in what we call the mapping step. This mapping step, which converts the resources into a common format that allows latter the merging, is usually per- formed with huge manual effort and thus makes the whole process very costly. Thus, we pro- pose a method to perform this mapping fully automatically. To test our method, we have ad- dressed the merging of two verb subcategoriza- tion frame lexica for Spanish, The results achieved, that almost replicate human work, demonstrate the feasibility of the approach. 1 Introduction The production, updating, tuning and maintenance of Language Resources for Natural Language Processing is currently being considered as one of the most promising areas of advances for the full deployment of Language Technologies. The reason is that these resources that describe, in one way or another, the characteristics of a particular language are necessary for Language Technologies to work. Although the re-use of existing resources such as WordNet (Fellbaum, 1998) in different applica- tions has been a well known and successful case, it is not very frequent. The different technology or application requirements, or even the ignorance about the existence of other resources, has pro- voked the proliferation of different, unrelated re- sources that, if merged, could constitute a richer repository of information augmenting the number of potential uses. This is especially important for under-resourced languages, which normally suffer from the lack of broad coverage resources. The research reported in this paper was done in the context of the creation of a gold-standard for sub- categorization frames of Spanish verbs to be used in lexical acquisition (Korhonen, 2002). We wanted to merge two hand-written, large scale Spanish lexica to obtain a new one that is richer and validated. Because subcategorization frames contain highly structured information, it was con- sidered a good scenario for testing new lexical resource merging methods. Several attempts at resource merging have been addressed and reported in the literature. Teufel (1995) and Chan & Wu (1999) were concerned with the merging of several source lexica for PoS tagging. The merging of more complex lexica has been addressed by Crouch and King (2005) who produced a Unified Lexicon with lexical entries for verbs based on their syntactic subcategorization in combination with their meaning, as described by WordNet, Cyc (Lenat, 1995) …

Content truncated to 500 words