Paper Type
Complete
Abstract
The aim of this study is to develop and validate an embedding-driven framework that automates the identification of research gaps in Systematic Literature Reviews (SLRs) by combining transformer-based keyword clustering with multi-criteria peripherality scoring to detect semantic peripheries. The workflow ranks articles by cosine distance, extracts and encodes keywords, clusters them with k-means in a semantic space, and scores gaps using a formula combining z-score peripherality, TF-IDF, rarity, and centroid distance. In a case study on gamification in marketing (14,302 records), it identified 25 research gaps and showed greater semantic coherence than VOSviewer’s co-occurrence map. Validation with five LLM annotators resulted in nearly 2x increase in GAP classification and a 30.4% rise in novelty score. The method can semi-automate gap identification, reducing processing time.
Paper Number
1695
Recommended Citation
Frankowski, Pawel Karol; Wiśniewska, Joanna; and Matysik, Sebastian, "Semantic Periphery Detection in Academic Keyword Space: An Embedding-Driven Framework for Automated Research Gap Identification" (2026). AMCIS 2026 Proceedings. 7.
https://aisel.aisnet.org/amcis2026/ai_aiaa/ai_aiaa/7
Semantic Periphery Detection in Academic Keyword Space: An Embedding-Driven Framework for Automated Research Gap Identification
The aim of this study is to develop and validate an embedding-driven framework that automates the identification of research gaps in Systematic Literature Reviews (SLRs) by combining transformer-based keyword clustering with multi-criteria peripherality scoring to detect semantic peripheries. The workflow ranks articles by cosine distance, extracts and encodes keywords, clusters them with k-means in a semantic space, and scores gaps using a formula combining z-score peripherality, TF-IDF, rarity, and centroid distance. In a case study on gamification in marketing (14,302 records), it identified 25 research gaps and showed greater semantic coherence than VOSviewer’s co-occurrence map. Validation with five LLM annotators resulted in nearly 2x increase in GAP classification and a 30.4% rise in novelty score. The method can semi-automate gap identification, reducing processing time.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIG AIAA