Improving Metaphor Detection with Frame Embedding Learning and Effective Context Denoising

Event details

This event will discuss language processing related issues and will have some implications in social sciences.

Abstract

Metaphor is a pervasive linguistic device, which attracts attention from both the fields of psycholinguistics and computational linguistics due to the key role it plays in the cognitive and communicative functions of language. However, metaphor detection is a challenging task as it requires understanding the nuanced relationships between abstract concepts embodied by the metaphoric expression and its surrounding context.

This talk will introduce two recent works that tackle metaphor detection. The first work proposes FrameBERT, a RoBERTa-based model that can explicitly learn and incorporate FrameNet Embeddings for concept-level metaphor detection. The second work introduces a flat, target-oriented tree structure that enables the model to effectively concentrate on the semantically relevant context of a target word, leading to enhanced metaphor detection.

Biography

Chenghua Lin is Professor of Natural Language Processing in the Computer Science Department, University of Manchester. His research interests are centred around Natural Language Processing and Generation (NLP/NLG), Machine Learning, and the intersection between them. He has published over 100 publications in the aforementioned areas and has received several awards for his research, including the Test of Time Award in CIKM 2020, the Best Paper Award Runner-up in INLG 2019, and the Best Paper Award in NLDB 2017.

He is currently Secretary of the ACL SIGGEN Board and Committee Member of the IEEE Speech and Language Processing Technical Committee. He has served in numerous program and (senior) area chairing roles for *ACL conferences, including publication co-chair for ACL 2023, workshop co-chair for AACL- IJCNLP 2022, and program co-chair for INLG 2019. He also serves as Associate Editor for Neurocomputing and a standing reviewer for TACL and the Computational Linguistics Journal

Back to events