Convergence of DB and IR

Attended a panel discussion this morning at a conference for database researchers. Quite an interesting panel discussion, titled "DB/IR - What Wine, What Bottle?", with panelists from Australia, India, USA, and Singapore. The focus of the discussion on what are the new directions of research that can be made in combining database and information retrieval techniques. The general flow of the discussion was on what constitute "Old Wine/Old Bottle" (i.e. research topics already well-studied) vs "New Wine/New Bottle" (i.e. forward-looking areas of research). Some of the key ideas brought about by the various panelists were:

One commented that some existing commercial database engines were not able to deal with the predicates <= 5 and < 6 properly, and return different selectivity estimates.

Others promising areas include:
- Core/basic database engine research.
- Video retrieval techniques
- Geo-enabled search
- P2P technology + IR
- Music information management (retrieval of the top K similar songs)
- Small footprints DBMS

Will post more about the other inspiring keynote speeches later once I get my notes organized.

My own personal lists of DB/IR topics
1) Question-Answering (QA) has been the focus of many IR researchers in the past few years, with a lot focussing on competing at the various TREC QA tasks. Can we apply similar QA techniques in answering database queries?

2) Would search engine benefit from the query-optimize-execute paradigm? Are query optimizers relevant in a search engine context? One of the advantages of using a declarative query language like SQL is that we are concerned with what kind of results we want, but we leave the tasks of finding the most efficient query plans to the query optimizers. Is the current query processing techniques for search engine optimal? Can we do better with a search engine optimizer? What is a search engine optimizer?