Vector Search: AI at the Service of Meaning

The thesaurus manages manual rules (A = B), while vector search uses AI to automatically deduce the semantic relationship between terms. AI offers significantly superior scale and precision for search thanks to vector data and numerical representations.

Yes. The Artificial Intelligence models used for vector search are often trained on vast multilingual corpora, allowing them to understand and match meaning regardless of the language. These capabilities rely on LLM language models, which are adapted to multilingual data sets.

No, it complements it. Vector search is added to the existing index (keywords, attributes) to improve relevance. Classic search handles precise filters, while AI—including generative AI—manages the understanding of meaning and the links between products via recommendation systems.

The AI engine is trained on general data, but it adapts over time based on interactions with your catalog’s results. This approach relies on varied data sets and can leverage open source models to boost performance.

The technical complexity is managed by the ElasticSuite/Gally tool. You only need a well-structured catalog; the AI infrastructure and vector management are handled by the platform, including the numerical representations required for processing.