Identifying core themes and research trends in SME innovation capabilities through topic modeling

Research on innovation in small and medium-sized enterprises (SMEs) has traditionally focused on verifying statistical relationships among isolated indicators, leaving the latent structure of innovation capabilities as an integrated construct largely unexplored. Unlike prior studies that rely on predefined theoretical frameworks or hypothesis-driven approaches, this study aims to address this gap by applying Latent Dirichlet Allocation (LDA) topic modeling to 3,590 academic abstracts retrieved from the Web of Science database, where the optimal number of topics was determined through coherence score and perplexity analysis, extracting latent thematic structures in a data-driven manner. The findings reveal four core topics of SME innovation capabilities: Entrepreneurial Competencies, Infrastructure, Activity, and Input. The proposed framework offers a new perspective on how SME innovation capabilities may be conceptualised as an integrated construct, providing empirically grounded insights for both policymakers and practitioners.