Towards a precise understanding of social entrepreneurship: An integrated bibliometric–machine learning based review and research agenda

This paper focuses on building a more precise and comprehensive understanding of the state of social entrepreneurship (SE) research by using an integrated bibliometric and unsupervised machine learning approach. Bibliometric analysis, along with Latent Dirichlet Allocation (LDA) for topic modeling enables us to identify key trends and themes in the SE domain. This approach is superior to tools and methods used in the past, which primarily employed systematic literature reviews and bibliometric analysis. While systematic manual literature reviews become impractical as the literature grows, bibliometric analysis focuses on the most cited articles, ignoring recent influential work, suffering from citation biases, and giving more weight to impact over thematic discovery. The methodology used by us overcomes these issues by first extracting large amounts of information through advanced computational methods and then using unsupervised machine learning to discover the latent themes and topics in this large collection of publications. This research uses the Scopus and Web of Science (WoS) databases to extract corpora of 3844 texts (titles, abstracts, and keywords) from published research on SE. We decipher the key trends in the literature and segregate them into three broad categories – individual attributes and motivation, organizational actions, and institutional conditions and development, with 21 sub-topics to enhance the understanding of this field of inquiry. This study is the first in the entrepreneurship domain to use this integrated approach to review the literature, and the findings lay the groundwork for future research.