Dongqing Zhu

1. Electronic Health Record Retrieval

The increasing prevalence of electronic health records (EHR) containing rich information about a patient's health and physical condition has the potential to transform research in health and medicine. In this project, we explore novel retrieval models as well as medical domain knowledge for building an EHR search system that can help medical researchers effectively identify cohorts for clinical studies.

Our system ranked 1st among automatic systems in TREC 2012 Medical Records Track [slides for TREC plenary talk]. The system was developed based on the research outcomes presented in the following publications:

D. Zhu and B. Carterette, Combining Multi-level Evidence for Medical Record Retrieval, In Proceedings of the 2012 international workshop on Smart health and wellbeing (SHB'12). ACM, New York, NY, USA, 49-56. [paper] [slides]

D. Zhu and B. Carterette, Improving Health Records Search Using Multiple Query Expansion Collections, In Proceedings of IEEE International Conference on Bioinformatics and Biomedicine (BIBM'12), 2012. [paper] [slides]

D. Zhu and B. Carterette, Using Multiple External Collections for Query Expansion, In Proceedings of the 20th Text REtrieval Conference (TREC), 2011. [paper] [slides]

2. TREC Microblog Track

In this project, we explore ways to address a realtime search task in the twitter world, where the user wishes to see the most recent but relevant tweets to the query.

B. Carterette, K. Naveen, R. Ashwani, D. Zhu, Simple RankBased Filtering for Microblog Retrieval: Implications for Evaluation and Test Collections, In Proceedings of the 20th Text REtrieval Conference (TREC), 2011.

3. Collecting & Analyzing User Preferences on SERP Layout

In this project, we carried out a pilot study using Amazon's Mechanical Turk to collect preference judgments between pairs of full-page layouts including both search results and image results. Specifically, we analyze the behavior of assessors that participated in our study to identify some patterns that may be broadly indicative of unreliable assessments. We believe this analysis can inform future experimental design and analysis when using crowdsourced human judgments.

D. Zhu and B. Carterette, An Analysis of Assessor Behavior in Crowdsourced Preference Judgments, In Proceedings of the SIGIR'10 Workshop on Crowdsourcing for Search Evaluation (CSE'10), Pages 21-26, 2010. [paper] [slides] [crowdsourcing interface]

4. Intelligent Joint Search in Health and Biomedical Databases

Health records & biomedical scientific literature are stored in different databases with different metadata. Few integrated search tools are available for finding basic science research that applies to clinical cases, or clinical cases relevant to ongoing basic research. Thus, we aim to build a search tool that can help translate biomedical science research to clinical care.

Dongqing Zhu, Ben Carterette, and Cathy Wu, Intelligent Joint Search in Health and Biomedical Databases, University of Delaware Research Foundation (UDRF), 2010. [poster] [interface demo]