Open Information Extraction
Abstract
Traditionally, Information Extraction (IE) has focused on satisfying precise, narrow, pre-specified requests from small homogeneous corpora (e.g., extract the location and time of seminars from a set of announcements). Shifting to a new domain requires the user to name the target relations and to manually create new extraction rules or hand-tag new training examples. This manual labor scales linearly with the number of target relations. This proposal introduces Open IE, a new extraction paradigm where the system makes a single data-driven pass over its corpus and extracts a large set of relational tuples without requiring any human input. The proposal also introduces TextRunner, a fully implemented, highly scalable Open IE system where the tuples are assigned a probability and indexed to support efficient extraction and exploration via user queries. Open IE is a very recent research breakthrough funded, in part, by our previous ONR grant on "Semantic Tractability on the World Wide Web". Here, we propose to study its efficacy and extend it in some important ways.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 31, 2010
- Accession Number
- ADA538482
Entities
People
- Oren Etzioni
Organizations
- University of Washington