Utilizing BERTopic Modeling for Concept Discovery in the Domain of Gerotranscendence and Solitude

Area of Science:

• Computational Linguistics
• Ontology Engineering
• Gerontology
• Psychology

Background:

• Traditional ontology development is manual, time-consuming, and prone to bias.
• Automated topic modeling offers a method to streamline concept discovery.
• This study explores BERTopic for extracting concepts from solitude and gerotranscendence literature.

Purpose of the Study:

• To apply BERTopic for semi-automated ontology development.
• To extract and refine candidate concepts for the PHASES Ontology.
• To compare default and custom BERTopic pipelines for topic modeling efficiency and quality.

Main Methods:

• Two BERTopic pipelines (default and custom with preprocessing) were implemented.
• Topic modeling was applied to corpora on solitude and gerotranscendence.
• Semantic similarity analysis (cosine similarity) was used to merge and integrate topics across domains.

Main Results:

• BERTopic identified 90 candidate ontology terms across both corpora.
• The custom pipeline yielded more fine-grained, domain-specific topics.
• Subject matter expert (SME) input was crucial for refining theory-laden concepts.

Conclusions:

• BERTopic offers an efficient, semi-automated method for ontology term extraction.
• Semantic similarity analysis facilitates cross-domain concept integration.
• This approach enhances ontology development and provides a replicable methodology.