NeuroCOLT workshop
on
Applications of Learning to Text and Images
Windsor, 30 April - 2 May 2001
Cumberland Lodge

Introduction to Multilingual Information Access
David Hull

dhull@whizbang.com

Download Presentation (tar file)

The internet now gives us unprecedented access to electronic information in many languages. However, finding and effectively exploiting information in different languages will require a new generation of multilingual information access (MLIA) technologies. Traditional techniques for information access (search, categorization, extraction, summarization, presentation) rely heavily on basic keyword matching strategies that are nearly worthless for detecting relationships between texts written in different languages. Therefore, the key issue in MLIA is finding a common representation for multilingual information via translation or concept-level matching. This tutorial will cover the following issues: (1) Language processing tools needed for MLIA (2) Overview of MLIA tasks with discussion on implications of multilinguality for each task (3) An introduction and comparison of different information access methods based on translation or concept-level matching. (4) Automatic creation of translation resources for MLIA using machine learning (5) MLIA problems in the Kermit project (group discussion)