My current main project is selective search. This project aims to alleviate the computational burden of small research institutions and start-up companies that need to perform search operations on web-scale indexes. By creating shards based on document similarity and performing judicious resource selection, computational costs of search can be reduced greatly by searching only the top k shards rather than the entire collection.
In a joint effort with Reyyan Yeniterzi, we participated in the ad-hoc search task of the Microblog Track of TREC 2012. The focus of our efforts was directed to addressing the vocabulary mismatch problem between the query and tweets. Two proposed solutions include query expansion through pseudo-relevance feedback and document expansion using URLs present in tweets. The resulting system was competitive and our best run placed in the top 10 of automatic runs.
The TREC paper is titled "Overcoming Vocabulary Limitations in Twitter Microblogs".
Previously, I have done research work in event detection, databases, and search engine internals. For more information, please see my curriculum vitae.