Move beyond literal matches.
Document collections are difficult to navigate when queries are short, ambiguous, or use different terms from the source material. Search quality depends on understanding intent and the structure of the text.
A search experience has to meet people in the language they actually use, even when it differs from the language of the documents they need.
Document collections are difficult to navigate when queries are short, ambiguous, or use different terms from the source material. Search quality depends on understanding intent and the structure of the text.
Kate worked on semantic search and text quality assessment, including query expansion and correction, word-sense disambiguation, topic modelling, classification, and services that made NLP components available to applications.
Dutch-language search and document analysis for government and legal information use cases, with backend APIs and demonstrations for product integration.
Combining language understanding with an interface and system that could help a user find relevant material. The right answer was not always a larger model; query handling and document quality also mattered.
Search and retrieval strategy, language-specific error analysis, data quality, and product thinking around how people find and judge information.