Clarifying User Intent in Enterprise Search

This was a 6-month internship where I did research on the clarification of user intent in search systems. My first few months were spent surveying clarification questions, query rewriting, and query expansion in natural language search. Subsequently, my work also focused on enterprise search systems that have a small search index, small user base, a lack of useful search log data, and domain-specific vocabulary. I developed a simple system of supplying search suggestions using key phrases within the document. I quantified some intuitively desirable characteristics of search suggestions in order to rank the key phrases- such as how well-formed, discriminative, and relevant they are. The original query used with our multi-word suggestions had higher query clarity and performance prediction scores. It was also able to perform better than the query-only setting in retrieval metrics but did not beat the query with single-word baseline setting.