THIS IS A CONTINUATION OF N00014-13-1-0023 Connecting the Dots in Huge Datasets:Structured,Personalized Outputs for Tackling Complex Information Needs
Abstract
Statement of Work:Develop efficient methods for finding relevant information in large document collections through expressive search that goes beyond keyword search, and methods for personalization of search results.Objective:Develop efficient methods for finding relevant information in large document collections through expressive search that goes beyond keyword search, and methods for personalization of search results.Approach:The PI s road map for achieving the objectives includes scientific developments along four major threads of research:(1) Beyond keyword search (discovering related documents): where a user can specify an information need as a structured collection of documents, and receive related documents that take into account a personalized notion of trust.(2) Connect the dots: where a user is able to examine a chain of events that connects two specified stories,personalized according to which topics the user deems important. (3) Automated issue maps: a structured, annotatedrepresentation of a topic that allows a user to examine the relevant relationships and arguments present in a specified collection of documents. (4) Fact checking the Web: where a user is able to investigate the veracity of a statement on the Web in situ, from many personalized points of view.These will be accomplished through developing structure graphs that reveal relationships in the document collection based on such concepts as influence, coherence, and similarity, which will be formalized through statistical approaches as well as natural language semantics. They will also develop rich user interaction methods that will personalize search results through feedback and learning.Overall Merit and ONR Mission/Relevance:This work is expected to advance our understanding of how to build methods for finding relevant information through searches that are expressive. This work is expected to deal with information overload confronting analysts, as well as automated systems, for findingrelevant information efficiently. It addresses primarily the Information Dominance focus area, as well as Autonomyfocus area.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 23, 2016
- Source ID
- N000141612795
Entities
People
- Carlos Guestrin
Organizations
- Office of Naval Research
- United States Navy
- University of Washington