CS Colloquium: Protiva Rahman (Ohio State University) - Amplifying Domain Expertise in Data Pipelines
Tue, Mar 19, 2019 @ 04:00 PM - 05:20 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Protiva Rahman, Ohio State University
Talk Title: Amplifying Domain Expertise in Data Pipelines
Series: Computer Science Colloquium
Abstract: Digitization of forms and electronic health records (EHR) has made data from diverse domains available for analysis. The specialized nature of the data require domain expert input at every step of the data analysis pipeline, including entry, cleaning, and analysis. Since domain experts (e.g. physicians) are highly skilled in their fields, their time is very valuable and expensive. Moreover, they often do not have any training in computer science or statistics, making it difficult for them to effectively interact with data. Thus, it is crucial that we make data interaction easy, efficient and effortless for experts. This involves amplifying or generalizing their inputs to multiple data points, reducing their time and effort.
In this talk, I will present Icarus, a system that leverages the database schema to amplify domain expert input during data cleaning. Icarus optimizes a weighted sum to guide the user to high-impact edits. Once a user fills in a cell, the system leverages the many-to-one relations in the database to suggest generalized update queries in the form of rules. These rules apply to a larger number of cells, amplifying the user\'s single edit.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Protiva Rahman is a fourth-year Ph.D. student in the Department of Computer Science and Engineering at the Ohio State University, advised by Professor Arnab Nandi. Her research interests include databases, human-computer interaction, visualization, and clinical informatics. Besides data cleaning, she has also worked on optimizing data entry interfaces for constrained interaction, guidelines for evaluating interactive systems and visualizations for domain expert consensus.
Host: Computer Science Department
Audiences: Everyone Is Invited
Contact: Computer Science Department