Entity Search Engine

As a part of dbms course, we had a guest lecture presented by our senior TA, Tao Cheng who is doing his Phd in UIUC. He explained about the concept of Entity Search engine and it was very interesting. Thought it would be useful to share my lecture notes with everyone.

Entity Search Engine : There is a need for a search engine to query various data entities which are available over the internet. For eg: Search the Web for Amazon.com's Customer Service Phone number

Current situation : Web Search engines serve only pages and not entities. Since they view the web as a set of documents and not entities, they donot support entity search. There is a need to take an entity view of the web and transform the traditional search pages into entities.

Solution – Entity Search Engine :
we look for specific entities like phone number or address or cheapest fare for a commodity etc.
The entity search engine addresses this exact problem. The search engine transforms the web documents into entities and the ranking system would rank the entities so that the results for queries could be ordered. Two of the interesting features that I noted at http://parrot.cs.uiuc.edu/entitysearch/ are

1. Ability to search specific patterns, say phone number of person X and the fax number of person X.
2. Filter the search to specific entities. For eg: Search the phone number of professors but filter it down to dept = database and univ = UIUC

Potential Issues I could see :
From a privacy perspective, I could see 2 potential problems

1. Information of certain high profile people would need to kept private
2. Certain sensitive entities like SSN should not be indexed or results displayed.

One solution is to have default filters for SSN / other sensitive search entities. For maintaing privacy of people information, we could maintain a "DONOT DISPLAY" list in which we could add the list of people who want to maintain privacy in the Internet world and then add this list to the filter while querying

Also, if a user tries to increase the false positive result by intentionally providing incorrect information over several resources on the web, then the entity search engine should be made intellegent enough to identify the false positive information. For eg: a user might have intentionally given his email id as smith@j.com in some site and would have given smith@q.com in some other site. It would be interesting to solution this problem in the entity search engine