Semantic Search: An Interview with Peter Mika, Yahoo! Research

Peter Mika is a researcher with Yahoo! Research in Barcelona, Spain. He will be in San Diego on June 19, 2010, to present a talk, The future face of Search is Semantic for Facebook, Google, and Yahoo!. In light of Peter’s talk, Barbara Starr — organizer of the San Diego Semantic Web meetup — asked him a few questions about semantic search.

What is Semantic Search?
Semantic Search, as it increasingly being used, is the idea of improving document search by using metadata or by searching on the metadata directly. An example of the first case is when we exploit metadata embedded in HTML — using microformats or Resource Description Framework (RDFa) — to improve some aspect of search. In the case of Yahoo! SearchMonkey, the metadata is used to enrich search result display. Examples of the second case are Semantic Web search engines such as Sindice that crawl and index metadata in RDF.

There is a growing interest in both areas due to the increasing amounts of metadata available, and because we expect that end users will still prefer to write keyword queries, instead of formulating their queries in RDF query languages such as SPARQL. In fact, an important task in Semantic Search is the interpretation of user queries.
As a user, how will Semantic Search improve my experience?
There are a number of end-user benefits. SearchMonkey demonstrates how the experience can benefit search-result presentation. It makes abstracts more informative by including data, images, and video.
Semantic Search will also bring entirely new functionality, such as providing direct answers to factual queries but also aggregating answers. As an example, for the query “san francisco concerts”, a search engine would be able to show a timeline of events taking place in san Francisco, as opposed to asking the user to visit 10 or more links and collect the data by herself.
Eventually, the search engine could replace all data aggregators and domain-specific (vertical) search engines. We are not quite there yet, but the potential is there.

How does it relate to social networks?
It’s orthogonal to social networks and social search. Something might be relevant to the user because of our own goals and interests, but that could be amplified by the goals, interests, recommendations, and experiences of our friends.
This also means that while search engines could benefit from access to a user’s social network, social networks could also benefit from metadata associated with Web pages. This has been recently demonstrated by Facebook’s Open Graph Protocol. OGP allows users to share and “like” objects (such as a movie) as opposed to Web pages. This kind of “semantic profiling” will eventually give Facebook a much better picture of what their users are interested in.