The Performance of Topic Evolution Based on a Feature Maximization Measurement for the Linguistics Domain

Abstract

Understanding the performance of the data mining approach and topic evolution in a certain scientific domain is imperative to capturing key domain developments and facilitating knowledge transfer within and across domains. Our research selects linguistics as an exploratory domain and exploits the feature maximization (FM) measurement for feature selection, combined with the contrast ratio to conduct the diachronic analysis for the linguistics domain’s topics. To accurately mine the linguistics domain’s topics and obtain the optimal clustering model selection, we exploit an integrated method associated with the deep embedding for clustering (DEC) algorithm based on the keywords-based Text Representation Matrix (KTRM) and Lamirel’s EC index and test the performance of this method. The results show that the FM measurement is applicable in the linguistics domain for topic mining, and the combinatory method has the advantage of an unbiased clustering optimization model and applies to the design of non-parameter clustering and algorithms from the low dimension to the high dimension of datasets. The findings suggest that this approach could be suitable for a diachronic analysis of topic evolution and facilitate the performance of topic detection. In addition, these findings of text detection can rise to knowledge fusion cognition with the factor of language as an available research objective in interdisciplinary research.

Keywords

• feature maximization
• <i>EC</i> index
• deep embedding clustering (DEC)
• linguistics domain
• topic evolution