Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{196581,
        author = {Satish Nishad and Abhijeet Kumar Ojha and Shubham and Karishma Samal and Hritik Tyagi},
        title = {A Comparative Experimental Study of Keyword Search vs. Semantic Search in MERN Stack Applications Using Vector Embeddings},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {3836-3845},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=196581},
        abstract = {This paper offers a statistically rigorous empirical comparison of keyword-based (TF-IDF style) search and cosine similarity vector embedding search within a MERN (MongoDB, Express.js, React, Node.js) stack environment. Unlike prior studies that rely on synthetic benchmarks or vendor-reported metrics, this research is based on a fully reproducible real-world experiment: a MongoDB Atlas cluster hosting 1,000 product documents with 128-dimensional hash-based embeddings, evaluated through 50 manually constructed test queries covering exact-match, synonym, and conceptual query types. Performance was measured using Precision@5 (P@5), Mean Reciprocal Rank (MRR), and Recall@10. Statistical significance was assessed with paired two-tailed t-tests and Cohen's d effect size. Results show that keyword search achieves significantly higher overall retrieval accuracy (MRR: 0.970 vs 0.432, t(49) = 7.719, p < 0.0001, d = 1.592), while semantic search outperforms keyword search on synonym-type queries (MRR: 0.938 vs 0.875) and is consistently 2.8 times faster (0.74ms vs 2.08ms, t(49) = 10.872, p < 0.0001, d = 2.172). These findings highlight that embedding model quality is the main factor influencing semantic search performance, and even lightweight hash-based embeddings offer measurable benefits for synonym-rich query workloads. The study presents a fully reproducible, open-source benchmark methodology for MERN developers, providing evidence-based recommendations for choosing search architecture in production deployments.},
        keywords = {Semantic Search, Keyword Search, Vector Embeddings, MERN Stack, MongoDB Atlas, Paired t-test, Cohen's d, Information Retrieval, Precision@K, MRR, Recall@K, Cosine Similarity, Node.js},
        month = {April},
        }