EmbedSLR: an open-source python framework for efficient embedding-based screening and bibliometric validation in systematic literature review

Efficient screening is essential for systematic literature reviews (SLRs), but traditional methods tend to be slow, biased, and error-prone. EmbedSLR improves this process by combining conventional bibliometrics with modern AI techniques. Its deterministic pipeline first ranks articles based on cosine distance between user queries and embeddings, then evaluates them using lightweight bibliometric indices. This approach outperformed keyword searches by increasing shared reference indicators 41.02 times and shared keywords 6.95 times in the presented case. Compatible with local environments and Google Colab, EmbedSLR supports natural language queries, making it accessible for researchers without programming skills while ensuring reproducible, high-quality SLRs. Our contribution is a deterministic, reproducible software pipeline that standardizes embedding‑based screening (cosine) and automatic bibliometric audit. The contribution of this research is a deterministic, repeatable software process that standardizes embedding-based (cosine) selection and automatic bibliometric control.