Grant: Collaborative Research and Development Grants

Methods for finding information in massive amounts of text are well established, and run in the background of familiar Web search engines such as Google. These methods enable anyone to find the needle in the haystack. The next challenge is to find a way to help experts and analysts in a given field to tap the tacit knowledge hidden in masses of text; not just to find the needle, but to understand how it relates to other complex issues, to make sense of the data on a higher level, and to reveal what needles they should be looking for in the first place. The proposed research aims to bring together text mining, text visualization and human computer interaction to create computer tools to support the sense-making activity of the human user. Key challenges include: the extraction and visualization of concepts, names and relations from large noisy text corpora; visualization of relations between concepts in text as graph structures; support for real-time visualization and interaction, which requires a careful trade-off between off-line and on-line processing; novel text visualization techniques and interaction techniques that permit the domain analyst to browse through the knowledge content of the text corpus and fine tune the text mining, without becoming a text mining expert. The industrial partner, AeroInfo Systems (Boeing Canada Operations Ltd.) has a major interest in the application of visual text analytics to aviation safety related corpora aiming to improve all aspects of safety in aircraft manufacturing, operation and maintenance. This is the main driver of the proposed project. Additional domains that will be addressed, to broaden the scope of the research, and the generality of the proposed methodologies, include: help desk applications (trouble ticket mining); community question-answering forum mining; research literature and patent mining; financial report and time series mining.