Analyzing the generalizability of the network-based topic emergence identification method

Tracking #: 2862-4076

This paper is currently under review
Sukhwan Jung
Aviv Segev

Responsible editor: 
Guest Editors DeepL4KGs 2021

Submission type: 
Full Paper
The field of topic evolution helps the understanding of the current research topics and their histories by automatically modeling and detecting the set of shared research fields in the academic papers as topics. This paper provides a generalized analysis of the topic evolution method for predicting the emergence of new topics, which can operate on any dataset where the topics are defined as the relationships of its neighborhoods in the past by extrapolating to the future topics. Twenty sample topic networks were built with various fields-of-study keywords as seeds, covering domains such as business, materials, diseases, and computer sci-ence from the Microsoft Academic Graph dataset. The binary classifier was trained for each topic network using 15 structural features of emerging and existing topics and consistently resulted in accuracy and F1 over 0.91 for all twenty datasets over the periods of 2000 to 2019. Feature selection showed that the models retained most of the performance using only one-third of the features used. Incremental learning was tested within the same topic over time and between different topics, which resulted in slight performance improvements in both cases. This indicates there is an underlying pattern to the neighbors of new topics com-mon to research domains, likely beyond the sample topics used in the experiment. The result showed the network-based new topic prediction can be applied to various research domains with different research patterns.
Full PDF Version: 
Under Review