Description
Generic
Rranking
Algebra for
Full
Text
GRAFT pairs (or grafts together) a relational algebra like language for expressing full-text queries that have complex word-position predicates with a generic scoring algebra capable of expressing many relevant scoring functions.
The key to making GRAFT work is a scoring-aware optimizer that considers both the chosen query and chosen scoring algebra implementation when optimizing a query. In this way the same optimizer may be used for many different scoring functions, including scoring functions that are defined after the optimizer is built.
People
- Alin Deutsch (Associate Professor - UC San Diego)
- Vasilis Vassalos (Associate Professor - Athens University of Economics and Business)
- Nathan Bales (PhD Candidate - UC San Diego)