Tuesday, 1st April 2014
Knowledge Base Completion via Search-Based Question Answering
Source: Research at Google
Over the past few years, massive amounts of world knowledge have been accumulated in publicly available knowledge bases, such as Freebase, NELL, and YAGO. Yet despite their seemingly huge size, these knowledge bases are greatly incomplete. For example, over 70% of people included in Freebase have no known place of birth, and 99% have no known ethnicity. In this paper, we propose a way to leverage existing Web-search–based question-answering technology to fill in the gaps in knowledge bases in a targeted way. In particular, for each entity attribute, we learn the best set of queries to ask, such that the answer snippets returned by the search engine are most likely to contain the correct value for that attribute.
+ Direct link to document (PDF; 554 KB)
Having begun his career in academic libraries, Adrian Janes has subsequently worked extensively in public libraries, chiefly in enquiry work as an Information Services librarian. In this role he has had particular responsibility for information from both the UK Government and the European Union. He wrote a detailed report on sources for the latter which was published by FreePint in 2007, and has contributed articles to FreePint and ResourceShelf. He is involved in training in information literacy and the use of online reference resources.
A Contributing Editor to DocuTicker, he also write reviews for Pennyblackmusic.
Adrian can be reached at firstname.lastname@example.org
More articles by Adrian Janes »
Please note: DocuTicker's editors collect citations for full-text PDF reports freely available on the web but we do not archive these reports. When you click a link to find and/or download the report, you are leaving the DocuTicker site. DocuTicker makes no representations regarding the ongoing availability of any report or any external resource. Links were accurate as of the date of posting.