BIPRE: Bi-directional Inter-Personal Relationship Extraction Task

Autor:	Pengyuan Liu, Yongsheng Tian (田永胜), Bingjie Du (杜冰洁)
Rok vydání:	2021
Předmět:	Interpersonal relationship Computer science business.industry Personal relationship Question answering Baseline model Artificial intelligence computer.software_genre business computer Relationship extraction Natural language processing Task (project management)
Zdroj:	Lecture Notes in Computer Science ISBN: 9783030811969 CLSW
DOI:	10.1007/978-3-030-81197-6_54
Popis:	The existing inter-personal relationship extraction task (the IPRE task) is a one-way task, and it does not take relationship in the opposite direction into consideration. For example, the IPRE task can extract the relationship in “A is the father of B”, but in such a circumstance, one still does not know whether the relationship between B to A is a son-father relationship or a daughter-father relationship. For many natural language processing tasks, it is important to identify specific and clear bi-directional relationships, especially in dialogue and question answering systems. Hence, this paper proposes a Bi-directional InterPersonal Relationship Extraction Task (BIPRE), establishes a corresponding dataset, and implements a Bert-based relationship extraction baseline model. Also, we propose a new model which combines the position and gender information of the persons. The experimental results show that our new model is effective and there is still much room for improvement in this task.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::a4a664fda080fa0b2c06753abe773ea4 https://doi.org/10.1007/978-3-030-81197-6_54 Zobrazit plný text záznamu