Enabling Structured Queries over Unstructured Documents

Abstract

With the information explosion on the Internet, finding precise answers efficiently is a prevalent requirement by many users. Today, search engines answer keyword queries with a ranked list of documents. Users might not be always willing to read the top ranked documents in order to satisfy their information need. It would save lots of time and efforts if the the answer to a query can be provided directly, instead of a link to a document which might contain the answer. To realize this functionality, users must be able to define their information needs precisely, e.g., by using structured queries, and, on the other hand, the system must be able to extract information from unstructured text documents to answer these queries. To this end, we introduce a system which supports structured queries over unstructured text documents, aiming at finding structured answers to the users' information need. Our goal is to extract answers from unstructured natural text, by applying various efficient techniques that allow fast query processing over text documents from the web or other heterogeneous sources. A key feature of our approach is that it does not require any upfront integration efforts such as the definition of a common data model or ontology.