Learning to Learn to Disambiguate: Meta-Learning for Few-Shot Word Sense Disambiguation

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 4517–4533 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 4517 Learning to Learn to Disambiguate: Meta-Learning for Few-Shot Word Sense Disambiguation Nithin Holla ILLC, University of Amsterdam nithin.holla7@gmail.com Pushkar Mishra Facebook AI pushkarmishra@fb.com Helen Yannakoudakis Dept. of Informatics, King’s College London helen.yannakoudakis@kcl.ac.uk Ekaterina Shutova ILLC, University of Amsterdam e.shutova@uva.nl Abstract The success of deep learning methods hinges on the availability of large training datasets an- notated for the task of interest. In contrast to human intelligence, these methods lack versa- tility and struggle to learn and adapt quickly to new tasks, where labeled data is scarce. Meta- learning aims to solve this problem by training a model on a large number of few-shot tasks, with an objective to learn new tasks quickly from a small number of examples. In this pa- per, we propose a meta-learning framework for few-shot word sense disambiguation (WSD), where the goal is to learn to disambiguate un- seen words from only a few labeled instances. Meta-learning approaches have so far been typ- ically tested in an N-way, K-shot classifica- tion setting where each task has N classes with K examples per class. Owing to its nature, WSD deviates from this controlled setup and requires the models to handle a large number of highly unbalanced classes. We extend sev- eral popular meta-learning approaches to this scenario, and analyze their strengths and weak- nesses in this new challenging setting. 1 Introduction Natural language is inherently ambiguous, with many words having a range of possible meanings. Word sense disambiguation (WSD) is a core task in natural language understanding, where the goal is to associate words with their correct contextual meaning from a pre-defined sense inventory. WSD has been shown to improve downstream tasks such as machine translation (Chan et al., 2007) and infor- mation retrieval (Zhong and Ng, 2012). However, it is considered an AI-complete problem (Navigli, 2009) – it requires an intricate understanding of language as well as real-world knowledge. Approaches to WSD typically rely on (semi-) supervised learning (Zhong and Ng, 2010; Mela- mud et al., 2016; K˚ageb¨ack and Salomonsson, 2016; Yuan et al., 2016) or are knowledge-based (Lesk, 1986; Agirre et al., 2014; Moro et al., 2014). While supervised methods generally outperform the knowledge-based ones (Raganato et al., 2017a), they require data manually annotated with word senses, which are expensive to produce at a large scale. These methods also tend to learn a classifica- tion model for each word independently, and hence may perform poorly on words that have a limited amount of annotated data. Yet, alternatives that in- volve a single supervised model for all words (Ra- ganato et al., 2017b) still do not adequately solve the problem for rare words (Kumar et al., 2019). Humans, on the other hand, have a remarkable ability to learn from just a handful of examples (Lake et al., 2015). This inspired researchers to investigate techniques that would enable machine learning models to do the same. …

Content truncated to 500 words