Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks, Nils Reimers and Iryna Gurevych, 2019Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) (Association for Computational Linguistics (ACL))DOI: 10.18653/v1/D19-1410 - This paper introduces Sentence-BERT, a method for generating semantically meaningful sentence embeddings that are highly effective for cosine-similarity based semantic search, a core technique in RAG systems.
Introduction to Information Retrieval, Christopher D. Manning, Prabhakar Raghavan, and Hinrich Schütze, 2008 (Cambridge University Press) - A fundamental textbook that provides a thorough explanation of vector space models, cosine similarity, and the theoretical basis of information retrieval, directly relevant to semantic search. Chapter 6 is particularly valuable.
Approximate Nearest Neighbor Search, Alexandr Andoni, Piotr Indyk, Thijs Laarhoven, Ilya Razenshteyn, Ludwig Schmidt, 2021Communications of the ACM, Vol. 64 (Association for Computing Machinery)DOI: 10.1145/3472066 - This survey article offers a comprehensive overview of approximate nearest neighbor (ANN) search algorithms, which are crucial for efficiently scaling similarity search in large-scale vector databases.
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks, Patrick Lewis, Ethan Perez, Aleksa Gordić, Vladimir Karpukhin, Katja Filippova, Marc'Aurelio Ranzato, Mihai S. Rădulescu, Mike Lewis, Yann N. Dauphin, and Douwe Kiela, 2020Advances in Neural Information Processing Systems (NeurIPS) (NeurIPS Foundation)DOI: 10.5591/neurips.2020.00949 - The foundational paper introducing the Retrieval-Augmented Generation (RAG) framework, detailing how the retrieval component, which relies on similarity search, is integrated with large language models for enhanced performance.