Hierarchical topic tree: A hybrid model comprising network analysis and density peak search
- Publisher:
- INT SOC SCIENTOMETRICS & INFORMETRICS-ISSI
- Publication Type:
- Conference Proceeding
- Citation:
- 18th International Conference on Scientometrics and Informetrics, ISSI 2021, 2021, pp. 1241-1252
- Issue Date:
- 2021-01-01
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Topic hierarchies can help researchers to develop a quick and concise understanding of the main themes and concepts in a field of interest. This is especially useful for newcomers to a field or those with a passing need for basic knowledge of a research landscape. Yet, despite a plethora of studies into hierarchical topic identification, there still lacks a model that is comprehensive enough or adaptive enough to extract the topics from a corpus, deal with the concepts shared by multiple topics, arrange the topics in a hierarchy, and give each topic an appropriate name. Hence, this paper presents a one-stop framework for generating fully-conceptualized hierarchical topic trees. First, we generate a co-occurrence network based on key terms extracted from a corpus of documents. Then a density peak search algorithm is developed and applied to identify the core topic terms, which are subsequently used as topic labels. An overlapping community allocation algorithm follows to detect topics and possible overlaps between them. Lastly, the density peak search and overlapping community allocation algorithms run recursively to structure the topics into a hierarchical tree. The feasibility, reliability, and extensibility of the proposed framework are demonstrated through a case study on the field of computer science.
Please use this identifier to cite or link to this item: