gensai.aiAn embedding is the numerical representation of a piece of content (word, sentence, document, image) as a vector that captures its meaning: two semantically similar contents have similar vectors. It is the bridge between language and computation, underlying semantic search and RAG.
In practice at Gensai
Gensai vectorises its clients' documents to enable meaning-based search in its answer engines.
How to choose your embeddings?
By testing several models on your own documents and real questions, in your users' language: a model that performs well in English does not necessarily do so in French. Document chunking, vector size and computing cost matter as much as the model itself. Open models make it possible to keep everything in-house.
Common mistake
Changing embedding model without recomputing everything. Vectors from two different models are not comparable: switching models means re-vectorising the entire base.
Same theme
Definition written with AI assistance, checked by Sébastien Lecocq, co-founder of Gensai.
Going beyond the definition?
From concept to project: Gensai builds custom, sovereign, GDPR-compliant AI solutions.
Related services: Chatbots and answer engines