Client
The challenge
- While preparing for a court case, lawyers analyze past judicial lines of reasoning relevant to the case, which involves manually searching and reviewing large volumes of data, consuming significant time.
What we did
- We developed a language model specialized for Polish legal language and fine-tuned it for legal text semantic similarity tasks.
- The model was utilized to build a highly effective hybrid search engine capable of scaling across more than 400,000 complex legal documents.
Benefits for the client
- The solution delivered search results that were more than twice as accurate as the traditional keyword-based search system previously used by our customer.
- Increased user satisfaction and engagement