Semantic Similarity Analysis of Theses Using Sentence Embeddings and Knowledge Graphs

Cusi I.J.D.P.; Mamani L.C.; Zuñiga-Rojas G.; Alvarez-Mamani E.; Vera-Olivera H. · Communications in Computer and Information Science (2026) · DOI: 10.1007/978-3-032-20322-9_7

Resumen

Theses authored by undergraduate students at Spanish-speaking universities are often omitted from major academic databases such as Scopus and Web of Science. These platforms prioritize English-language publications, including indexed journal articles and papers published in the proceedings of major conferences. As a result, these works, often written in Spanish and stored in institutional repositories, are underrepresented in scientific information systems, making the review of the state of the art a challenging task, particularly when aiming to include local and undergraduate academic contributions. This work presents a Knowledge Graph (KG) constructed from 7739 thesis authored by undergraduate students at the Universidad Nacional de San Antonio Abad del Cusco (UNSAAC), Perú. The documents, written in Spanish, were represented using vector embeddings generated by pretrained language models. Semantic similarity between documents was computed using cosine distance. Based on these similarities, a KG was built and modeled in Neo4j. The KG was further enriched with institutional metadata, including faculty and academic programs. This structure allows for the identification of conceptual links among documents and the detection of thematic communities within the collection. The source code, datasets, and implementation details are available at: https://github.com/JanelaCusi-03/semantic-thesis-similarity-KG. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

Instituciones: Universidad Peruana de Ciencias Aplicadas; Universidad Nacional de San Antonio Abad del Cusco; Universidad Andina del Cusco