Dynamic Clustering Combination for Concept Inference in Online Messages

Abstract

In the learning process of language and knowledge the construction and inference of the relations of concepts are very important steps This paper is motivated by observing the inferences process from that children have their own few and known conceptual vocabulary to generate unknown concepts Then three aspects are included to explore this conceptual vocabulary inference module They are (1) the feasibility of concept extraction (2) the appropriateness of concept inference methods and (3) the capturing of concept variations This research has conducted a concept vocabulary inference system and inference logic correspondence method to achieve the construction of a variety of conceptual vocabulary similar to people's association extension and jumping of conceptual vocabulary by a computational approach In experimental results the average performance of concept inference of the proposed Dynamic Clustering Combination and the retrieval-based combination approaches are better than the ones of general clustering methods Moreover our proposed approach can reach the practical diversity of concept extension

Date of Award	2020
Original language	English
Supervisor	Wei-Guang Teng (Supervisor)