Research on the Construction of Non-subject-object Argument Knowledge Base of Chinese Verbs

• Published in: 2022 International Conference on Asian Language Processing (IALP) pp. 451 - 457
• Main Authors: Wang, Chengwen, Xun, Endong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 27.10.2022
• Subjects: argument role priorities; Dictionaries; Knowledge based systems; Labeling; Linguistics; Natural language processing; non-subject-object argument; Semantics; structure retrieval; Syntactics
• DOI: 10.1109/IALP57159.2022.9961281

Whether in the field of linguistics or natural language processing (NLP), the study of Chinese verb arguments has always been a priority. The knowledge base of verb arguments is an essential resource for linguistic research and natural language processing research. However, previous knowledge bases have the following problems: (1) The method of constructing argument knowledge with verbs as items only lists few instances that can act as verb-argument roles, and cannot comprehensively present a large number of instances that act as argument roles. (2) Regarding determining the number of argument roles dominated by verbs, except for subject and object arguments, there is no clear operational standard for how many non-subject-object argument roles a verb dominates and how to determine them. Therefore, we propose an argument knowledge acquisition strategy based on big data, relying on a structured corpus and a structured retrieval system to acquire verb-argument instances from a large-scale corpus. At the same time, a calculation method for automatically determining the priority of verb non-subject-object argument is proposed, and preliminary experiments verify its effectiveness.