L2F/INESC-ID at SemEval-2019 Task 2: Unsupervised Lexical Semantic Frame Induction using Contextualized Word Representations

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 130–136 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 130 L2F/INESC-ID at SemEval-2019 Task 2: Unsupervised Lexical Semantic Frame Induction using Contextualized Word Representations Eug´enio Ribeiro1,2, Vˆania Mendonc¸a1,2, Ricardo Ribeiro1,3, David Martins de Matos1,2, Alberto Sardinha1,2, Ana L´ucia Santos4,5, Lu´ısa Coheur1,2 1 INESC-ID Lisboa, Portugal 2 Instituto Superior T´ecnico, Universidade de Lisboa, Portugal 3 Instituto Universit´ario de Lisboa (ISCTE-IUL), Lisboa, Portugal 4 Centro de Lingu´ıstica da Universidade de Lisboa, Portugal 5 Faculdade de Letras da Universidade de Lisboa, Portugal eugenio.ribeiro@l2f.inesc-id.pt, vania.mendonca@tecnico.ulisboa.pt Abstract Building large datasets annotated with seman- tic information, such as FrameNet, is an ex- pensive process. Consequently, such resources are unavailable for many languages and spe- cific domains. This problem can be alleviated by using unsupervised approaches to induce the frames evoked by a collection of docu- ments. That is the objective of the second task of SemEval 2019, which comprises three sub- tasks: clustering of verbs that evoke the same frame and clustering of arguments into both frame-specific slots and semantic roles. We approach all the subtasks by applying a graph clustering algorithm on contextualized embedding representations of the verbs and ar- guments. Using such representations is appro- priate in the context of this task, since they provide cues for word-sense disambiguation. Thus, they can be used to identify different frames evoked by the same words. Using this approach we were able to outperform all of the baselines reported for the task on the test set in terms of Purity F1, as well as in terms of BCubed F1 in most cases. 1 Introduction The Frame Semantics theory of language (Fill- more, 1976) states that one cannot understand the meaning of a word without knowing the context surrounding it. That is, a word may evoke dif- ferent semantic frames depending on its context. Considering this relation, sets of frame definitions and annotated datasets that map text into the se- mantic frames it evokes are important resources for multiple Natural Language Processing (NLP) tasks (Shen and Lapata, 2007; Aharon et al., 2010; Das et al., 2014). The most prominent of such resources is the FrameNet (Baker et al., 1998), which provides a set of more than 1,200 generic semantic frames, as well as over 200,000 anno- tated sentences in English. However, this kind of resource is expensive and time-consuming to build, since both the definition of the frames and the annotation of sentences require expertise in the underlying knowledge. Furthermore, it is difficult to decide both the granularity and the domains to consider while defining the frames. Consequently, such resources only exist for a reduced amount of languages (Boas, 2009) and even English lacks domain-specific resources in multiple domains. The problem of building semantic frame re- sources can be alleviated by using unsupervised approaches to induce the frames evoked by a collection of documents. The second task of SemEval 2019 aims at comparing unsupervised frame induction systems for building semantic frame resources for verbs …

Content truncated to 500 words