mining concepts from text
Tutorial 1 at ICON 2009, to be held Monday, 14 December 2009, at IIIT Hyderabad, is proposed by Prof Sutanu Chakraborti, IIT Madras. The duration is half a day, 0930–1300 hours.
Abstract
The Knowledge Acquisition bottleneck is one of the biggest challenges faced by Natural Language Processing. In this tutorial we discuss issues in reducing knowledge acquisition overloads in the problem of mining concepts from textual documents. In the context of this tutorial, concepts are induced from natural tendencies of words to associate with each other in a corpus; this notion is more fluid than concepts in Information Extraction which are specific grammatical patterns. We first motivate the need for concept mining using a concrete application domain that deals with leveraging corporate repositories to aid humans in tasks such as authoring proposals or processing employee feedback. We show how introspective knowledge, viz. the knowledge that is mined based on the document collection alone, can be effectively supplemented with background and linguistic knowledge. We explore five different formalisms for concept representation, and examine in detail introspective algorithms for concept mining in each of these categories. A comparative analysis of these techniques, highlighting their competences and weaknesses across diverse problem needs and nuances, will be presented. We then discuss approaches to integrate knowledge from background knowledge as in Wikipedia, and linguistic knowledge as in WordNet. We also sketch techniques to exploit additional dimensions of text like link structure and the social networks from which they originate. Many applications benefit from concept visualization and the provision for interactive concept refinement; these will be briefly covered. We conclude by identifying significant research issues.
About the Presenter
Dr. Sutanu Chakraborti is an Assistant Professor at the Department of Computer Science and Engineering, Indian Institute of Technology, Madras (IITM). His research interests include Case-Based Reasoning, NLP and Text Mining, and Machine Learning. At IITM, he offers electives on Memory Based Reasoning and Natural Language Processing. Prior to joining IITM, he was with the Arificial Intelligence research group at Tata Research Development and Design Centre, Pune, where he has 7 years of industry R&D experience as researcher and project leader.
organizing bodies
latest updates
23rd nov 2009
The technical schedule has been released. 
23rd nov 2009
Click here for details of the panel discussion.
contact us
ICON-2009 Secretariat
Language Technologies Research Centre
International Institute of Information Technology
Gachibowli
Hyderabad - 500032, India
Tel: +91-40.23001412
Fax: +91-40.66531413
icon2009@iiit.ac.in