| ssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss |
Quality Aware Management
(Supported by: NSF CSR SMALL )(Collaborators: Microsoft Research, Oak Ridge National Labs, Morehouse College, NCWIT, OSU Office of Diversity and Inclusion )
Research Challenge:
For many next generation services, data will grow much faster than processing capacity. These services will access only a portion of the data relevant to a user's query. Our research helps services automatically 1) identify the most valuable data for each user request and 2) ensure it can be accessed readily and cost effectively. Thus far, we have focused on natural language processing services, e.g., search engines and question-answering systems. For a specific user request, NL data often contains redundancies, e.g., multiple web pages with the same information or multiple word phrases with same meaning. These redundancies can crowd out useful data causing to NLP services to produce poor quality results. Our contribution is systems support to measure quality for distributed services and, subsequently, manage resources cost effectively within quality constraints.
Recommended Reading:
Cache Provisioning for Interactive NLP Services, LADiS 2013
Balanced and Predictable Networked Storage, DCPerf 2013
Media and Press Release:
 |
Outreach Event: We recently brought middle school students participating in Camp Engineer to the ReRout Labs for a friendly Jeopardy match. The middle schools students competed againt our Open Ephyra system (Open Ephyra is a question answering system like IBM Watson). The Youtube video to the left summarizes the event and unveils the winner. |