Result-Driven Database Recommendations

Ymal Results

In our position paper in PersDB 2009, we consider extending relational database systems with a recommendation functionality. In particular, we propose that, along with the results of each query, the user gets additional recommended results that we call "You May Also Like" or Ymal results. We discusse a suite of different approaches to computing Ymal results and exploite one that uses only the database content and the query results. Some preliminary evaluation results are also provided.

  • Current-state techniques: Techniques that exploit the content and schema of a current query result and database instance.
  • History-based techniques: Techniques that exploit the history of previously submitted queries to the database system, e.g., by using query logs or logs of query results.
  • External sources techniques: Techniques that exploit resources external to the database, such as related published results and reports, relevant web pages, thesaurus or ontologies.

Related work:

Predicting interesting results:

OLAP recommendations:

Auto-completing SQL queries:

Facet search:

Collaborative ranking for streaming results:

The need for database exploration: