Managing Relevance Across Mixed Content Types Answer Sets: A Case Study
Posted: 16 Dec 2019
Date Written: December 16, 2019
At LexisNexis our products typically span a range of content types which are searched separately, returning discrete answer sets which are optimized for each content type. Using the UK’s Lexis®PSL as a case study, this talk looks at a different approach, which returns to the user a single result list spanning multiple content types.
The Lexis®PSL service enables users to search for and view research and practical guidance content associated with legal matters, as well as primary content such as cases and legislation if searched for specifically. A key feature of this service is the presentation of a single results list which spans research and practical guidance content types.
We first look at the make-up of a mixed content type answer set in the legal domain as well as the drivers behind this approach from the user’s perspective – i.e., why does it matter?
Having established the what and the why, we then share learnings from the Lexis®PSL experience. We will look at approaches and challenges to tuning relevance across a diverse answer set, which will include methods of measuring relevance quality particularly in the context of query intent.
Keywords: search relevance measurement, ranking algorithms, query intent
Suggested Citation: Suggested Citation