Model-Agnostic Meta-Learning for Relation Classification with Limited Supervision

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5873–5879 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 5873 Model-Agnostic Meta-Learning for Relation Classification with Limited Supervision Abiola Obamuyide Department of Computer Science University of Sheffield avobamuyide1@sheffield.ac.uk Andreas Vlachos Dept. of Computer Science and Technology University of Cambridge andreas.vlachos@cst.cam.ac.uk Abstract In this paper we frame the task of supervised relation classification as an instance of meta- learning. We propose a model-agnostic meta- learning protocol for training relation classi- fiers to achieve enhanced predictive perfor- mance in limited supervision settings. Dur- ing training, we aim to not only learn good parameters for classifying relations with suf- ficient supervision, but also learn model pa- rameters that can be fine-tuned to enhance predictive performance for relations with lim- ited supervision. In experiments conducted on two relation classification datasets, we demonstrate that the proposed meta-learning approach improves the predictive performance of two state-of-the-art supervised relation clas- sification models. 1 Introduction Relation classification, the task of determining the relationship that exists between two entities, is a long-standing challenge in artificial intelligence with many downstream applications, including question answering, knowledge base population and web search. A variety of supervised meth- ods have been proposed in the literature for this task (Zelenko et al., 2003; Bunescu and Mooney, 2005; Mintz et al., 2009; Surdeanu et al., 2012; Riedel et al., 2013). Current approaches are pre- dominantly supervised models based on neural networks, for instance recursive neural networks (Socher et al., 2012; Hashimoto et al., 2013), con- volutional neural networks (Zeng et al., 2014; Nguyen and Grishman, 2015), recurrent neural networks (Zhang and Wang, 2015; Xu et al., 2015; Zhang et al., 2017) or a combination of recur- rent and convolutional neural networks (Vu et al., 2016). The performance of these approaches re- lies mostly on the quantity of their training data. However, labelled training data can be expensive to obtain and available only in limited quantities. It is therefore pertinent to develop methods that re- duce their reliance on large quantities of labelled training data. In this work we propose a model-agnostic pro- tocol for training supervised relation classification systems to achieve higher predictive performance in limited supervision settings, motivated by the observation that meta-learning leads to learning a better parameter initialization for new tasks than ad hoc multi-task learning across all tasks (Finn et al., 2017). We show that relation classification can be approached from a meta-learning perspec- tive, and propose a model-agnostic meta-learning protocol for training relation classification mod- els that explicitly learns a model parameter ini- tialization for enhanced predictive performance across all relations with limited supervision. Dur- ing training, our algorithm considers all relations and their instances as coming from a joint distribu- tion, and seeks to learn model parameters that can be quickly adapted using each relation’s training instances to enhance predictive performance on its test set. In experiments on two relation classification datasets, we apply the proposed approach to two relation classification …

Content truncated to 500 words