University of Pittsburgh at TREC 2014 Microblog Track
Abstract
An ad hoc retrieval task aims at return the most relevant documents based on a particular query. And high precision and recall always depends on clear query and elaborative documents. If the query is ambiguous while document is short and general, common retrieval models may not work well on the feedback. In this way, more expansive information will be needed to add in order to implement original queries and documents. That is the main purpose of microblog track of 2014 TREC Conference. The paper describes the first participation of University of Pittsburgh in 2014 TREC microblog track. The data is based on tweet collection which gathered in 2013. We got two runs for the final results which are base on BM25 feedback algorithm. The details of our retrieval model include query expansion, document expansion and retrieval model for the final rank.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2014
- Accession Number
- ADA618725
Entities
People
- Rui Bi
- Zheng Gao
Organizations
- University of Pittsburgh