A Knowledge-Guided Framework for Frame Identification

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 5230–5240 August 1–6, 2021. ©2021 Association for Computational Linguistics 5230 A Knowledge-Guided Framework for Frame Identification Xuefeng Su1,2, Ru Li1,3,∗, Xiaoli Li4, Jeff Z.Pan5, Hu Zhang1, Qinghua Chai1, Xiaoqi Han1 1. School of Computer and Information Technology, Shanxi University, Taiyuan, China 2. School of E-commerce and Logistics, Shanxi Vocational University of Engineering Technology, Taiyuan, China 3. Key Laboratory of Computational Intelligence and Chinese Information Processing of Ministry of Education, Shanxi University, Taiyuan, China 4. Institute for Infocomm Research, A*Star, Singapore 5. ILCC, School of Informatics, University of Edinburgh, UK {suexf,xiaoqisev}@163.com, {liru,zhanghu,charles}@sxu.edu.cn xlli@ntu.edu.sg, j.z.pan@ed.ac.uk Abstract Frame Identification (FI) is a fundamental and challenging task in frame semantic parsing. The task aims to find the exact frame evoked by a target word in a given sentence. It is generally regarded as a classification task in existing work, where frames are treated as discrete labels or represented using one-hot embeddings. However, the valuable knowl- edge about frames is neglected. In this pa- per, we propose a Knowledge-Guided Frame Identification framework (KGFI) that inte- grates three types frame knowledge, including frame definitions, frame elements and frame- to frame relations, to learn better frame rep- resentation, which guides the KGFI to jointly map target words and frames into the same em- bedding space and subsequently identify the best frame by calculating the dot-product sim- ilarity scores between the target word embed- ding and all of the frame embeddings. The ex- tensive experimental results demonstrate KG- FI significantly outperforms the state-of-the- art methods on two benchmark datasets. 1 Introduction Frame Identification (FI) aims to find the exact frame evoked by a target word in a given sentence. A frame represents an event scenario, and possess- es frame elements (or semantic roles) that partici- pate in the event (Hermann et al., 2014), which is described in the FrameNet knowledge base (Bak- er et al., 1998; Ruppenhofer et al., 2016) ground- ed on the theory of Frame Semantics (Fillmore et al., 2002). The theory asserts that people under- stand the meaning of words largely by virtue of the frames which they evoke. In general, many words are polysemous and can evoke different frames in different contexts. As shown in Figure 1, the word stopped e- vokes the frame Activity stop and the frame ∗Corresponding author. The company stopped producing the profitable toy. Activity_stop stop.v Agent Activity Process_stop stop.v Process Time The fighting has stopped for more than two years . Figure 1: Two annotated examples with the target word marked in bold and frame elements (semantic roles) in rounded rectangles. The target word stopped (stop.v denotes its form of lexical unit) evokes the frame Activity stop and the frame Process stop respectively in different contexts. Here, the key to distinguish these two frames is identifying whether the subject (The com- pony or The fighting) of stopped is an Agent or a Pro- cess (see the …

Content truncated to 500 words