Measuring Fine-Grained Domain Relevance of Terms: A Hierarchical Core-Fringe Approach

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 3641–3651 August 1–6, 2021. ©2021 Association for Computational Linguistics 3641 Measuring Fine-Grained Domain Relevance of Terms: A Hierarchical Core-Fringe Approach Jie Huang1,3 Kevin Chen-Chuan Chang1,3 Jinjun Xiong2,3 Wen-mei Hwu1,3 1University of Illinois at Urbana-Champaign, USA 2IBM Thomas J. Watson Research Center, USA 3IBM-Illinois Center for Cognitive Computing Systems Research (C3SR), USA {jeffhj, kcchang, w-hwu}@illinois.edu jinjun@us.ibm.com Abstract We propose to measure fine-grained domain relevance– the degree that a term is relevant to a broad (e.g., computer science) or narrow (e.g., deep learning) domain. Such measure- ment is crucial for many downstream tasks in natural language processing. To handle long- tail terms, we build a core-anchored semantic graph, which uses core terms with rich descrip- tion information to bridge the vast remaining fringe terms semantically. To support a fine- grained domain without relying on a matching corpus for supervision, we develop hierarchi- cal core-fringe learning, which learns core and fringe terms jointly in a semi-supervised man- ner contextualized in the hierarchy of the do- main. To reduce expensive human efforts, we employ automatic annotation and hierarchi- cal positive-unlabeled learning. Our approach applies to big or small domains, covers head or tail terms, and requires little human effort. Extensive experiments demonstrate that our methods outperform strong baselines and even surpass professional human performance.1 1 Introduction With countless terms in human languages, no one can know all terms, especially those belonging to a technical domain. Even for domain experts, it is quite challenging to identify all terms in the do- mains they are specialized in. However, recogniz- ing and understanding domain-relevant terms is the basis to master domain knowledge. And having a sense of domains that terms are relevant to is an initial and crucial step for term understanding. In this paper, as our problem, we propose to measure fine-grained domain relevance, which is defined as the degree that a term is relevant to a 1The code and data, along with several term lists with domain relevance scores produced by our meth- ods are available at https://github.com/jeffhj/ domain-relevance. given domain, and the given domain can be broad or narrow– an important property of terms that has not been carefully studied before. E.g., deep learn- ing is a term relevant to the domains of computer science and, more specifically, machine learning, but not so much to others like database or compiler. Thus, it has a high domain relevance for the former domains but a low one for the latter. From another perspective, we propose to decouple extraction and evaluation in automatic term extraction that aims to extract domain-specific terms from texts (Amjadian et al., 2018; H¨atty et al., 2020). This decoupling setting is novel and useful because it is not limited to broad domains where a domain-specific corpus is available, and also does not require terms must appear in the corpus. A good command of domain relevance of terms will facilitate many …

Content truncated to 500 words