Survey on the Use of Typological Information in Natural Language Processing

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 1297–1308, Osaka, Japan, December 11-17 2016. Survey on the Use of Typological Information in Natural Language Processing Helen O’Horan1∗, Yevgeni Berzak2∗, Ivan Vuli´c1, Roi Reichart3, Anna Korhonen1 1 Language Technology Lab, DTAL, University of Cambridge 2 CSAIL MIT 3 Faculty of Industrial Engineering and Management, Technion, IIT helen.ohoran@gmail.com berzak@mit.edu iv250@cam.ac.uk roiri@ie.technion.ac.il alk23@cam.ac.uk Abstract In recent years linguistic typology, which classifies the world’s languages according to their functional and structural properties, has been widely used to support multilingual NLP. While the growing importance of typological information in supporting multilingual tasks has been recognised, no systematic survey of existing typological resources and their use in NLP has been published. This paper provides such a survey as well as discussion which we hope will both inform and inspire future work in the area. 1 Introduction One of the biggest research challenges in NLP is the huge global and linguistic disparity in the availability of NLP technology. Still, after decades of research, high quality NLP is only available for a small number of the thousands of languages in the world. Theoretically, we have two solutions to this problem: i) development of universal, language-independent models which are equally applicable to all natural language, regardless of language-specific variation; ii) comprehensive systematisation of all possible variation in different languages. The field of linguistic typology offers valuable resources for nearing both of these theoretical ideals: it studies and classifies world’s languages according to their structural and functional features, with the aim of explaining both the common properties and the structural diversity of languages. Many of the current popular solutions to multilingual NLP: transfer of information from resource-rich to resource-poor languages (Pad´o and Lapata, 2005; Khapra et al., 2011; Das and Petrov, 2011; T¨ackstr¨om et al., 2012, inter alia), joint multilingual learning (Snyder, 2010; Cohen et al., 2011; Navigli and Ponzetto, 2012, inter alia), and development of universal models (de Marneffe et al., 2014; Nivre et al., 2016, inter alia), either assume or explicitly make use of information related to linguistic typology. While previous work has recognised the role of linguistic typology (Bender, 2011), no systematic survey of typological information resources and their use in NLP to date has been published. Given the growing need for multilingual NLP and the increased use of typological information in recent work, such a survey would be highly valuable in guiding further development. This paper provides such a survey for structural typology, the areas of typological theory that consider morphosyntactic and phonological features1, which has been the main focus of typology research in both linguistics and NLP. We begin by introducing the field of linguistic typology and the main databases currently available (§ 2). We then discuss the role and potential of typological information in guiding multilingual NLP (§ 3). In § 4 we survey existing NLP work in terms of how typological information has been developed (4.1) and integrated in multilingual application tasks (4.2). § 5 discusses future …

Content truncated to 500 words