About the Founder
Founder’s Earlier Work · 02
Finding Useful Translation History
Semantic search and fuzzy text matching to surface useful prior translation segments.
Previous Work · Translation RetrievalThe Problem
A useful prior translation may express the same idea with different words. Exact text matching alone can miss it; semantic similarity alone can miss a close phrase or term.
My Role
I worked on retrieving previous translation segments using sentence-transformer embeddings and pgvector, combined with fuzzy text matching.
The Approach
- Represent segments as embeddings so searches can surface related meaning across wording differences.
- Use fuzzy text matching alongside semantic search to catch near-exact phrasing, names, and terminology.
Project Context
Prior project, completed by the founder before the company. This summary describes the retrieval approach without client data or measured accuracy claims.
Have a related problem to solve?
Discuss Your Project