Use case from prospect:
Say there are three titles in the data store:
- Beautiful and Beyond
- Beyond Beautiful
If search query term is “beautiful”, titles should always be weighted or scored to return in the order above, i.e. exact match to term scored highest, exact term at the beginning of the string next highest, exact term somewhere else in the string after that. Need a method to request this type of scoring added to the FTS service. For ties within these broad categories, standard tf/idf scoring and custom weighting could continue to apply.
After some discussion on the Slack channel we have reached the conclusion that most of the required elements required to do this computation are available already via the term vectors (locations-field-term pos, start, and end).