Fact extraction from organizational corpora

点击翻译按钮获取中文摘要

Organizational knowledge may be described in terms of facts, relationships and events. This information isoften expressed or shared within an organization in resources such as memos, financial statement, internal andexternal postings, bulletin boards, etc. It is not easy to search or process queries over such content. A first steptoward building a searchable repository of factual organizational knowledge is to effectively collect suchknowledge from large collections of textual documents. In this project, University of Alberta in partnershipwith DevFacto, a software technology and consulting company in Canada with a base in Edmonton, willexamine the feasibility and the effectiveness of extracting high quality facts and events from organizationalresources and memos. The work involves putting together a preliminary software system, based on theliterature and resources in public domain as well as those developed at the University of Alberta, and carry outexperiments to evaluate its performance in extracting facts and events from organizational resources, identifiedor provided by DevFacto. The developed system will feed data to Sparrow, a mobile application developed byDevFacto, to provide a better search of the organizational knowledge.Although the proposed project will mostly leverage existing algorithms from the literature, identifying andcombining the algorithms and evaluating them will be conducted in this project. Some of the algorithms mayalso be extended or revised to better work for the particular domain being studied. The project will beconducted with guidance from industry experts in DevFacto. The project will provide a training environmentfor graduate and undergrad students on topics related to information extraction, event detection and entityresolution.