From Levenshtein Distance to Neural Search: a Comprehensive Approach to Latent Semantic Search based on Functional Reference Books
Grigorii V. Orlov, A. N. Kalinichenko · 2025
The article discusses the use of various algorithms and methods to implement latent semantic search through reference data and functionality catalogs. Starting with classical methods such as Levenshtein distance and TF-IDF, the author delves deeper into modern approaches, including BM25, neural networks (DeepSeek), and large language model technologies (GigaChat). The developed approaches are aimed at improving the accuracy and relevance of the search by combining text processing methods, increasing the understanding of the user's query and the context in the text.