Copyright 2009 Digital Enterprise Research Institute. All rights reserved. Digital Enterprise Research Institute Expert (and Novice) Finding Within Shared Workspaces 03/02/2008 DERI Conference Room Peyman Nasirifard
Digital Enterprise Research Institute How do I do it? Live Demo! Challenges Conclusion and Future Work Questions Agenda
Digital Enterprise Research Institute What do we use? We do NOT use We do NOT use Wikipedia We do NOT use Google (co-occurence, etc.) We do NOT use DBLP We DO use Online Shared Workspaces – Currently BSCW We use two main elements / components of shared workspace Log file: Contains the transactions within shared workspace Document: Provides input for expertise extractor How Do I Do It?
Digital Enterprise Research Institute How Do I Do It? II Analyzing documents to extract key phrases Analyzing log files to see which persons did what events on what documents Rule of Thumb: A user is expert in topic X, if s/he has created or revised a document that contains topic X. A user is familiar with topic Y, if s/he has just read a document that contains topic Y
Digital Enterprise Research Institute Live Demo! Demo-based computing: Demos are always better than boring presentations!
Digital Enterprise Research Institute Challenges Ambiguity: “output parameter” is not really an expertise! There is no 100% accurate and automated Key phrase extraction algorithm from a document Solution: We had a semi-automated approach and we removed the terms with low confidences Future work: We will enable end users to remove inappropriate terms from their profiles and the algorithm can “learn” or “be trained” Organization Expertise profile vs. Person Expertise Profile We noticed that in some organizations, a single person tackles always with shared workspace and this may generate some noises in expertise profiles Solution: Building organization expertise profile besides person can address this issue Purifying Log files Some log records are noisy: They do not have the pre-assumed structure Solution: We defined some patterns and excluded the records that do not follow the patterns Expertise Granularity Rule of Thumb is not fine-grained enough Knud: Taking to account the version history of deliverables Similar phrases: “Semantic Web” and “SIOC” are from the same domain. This issue is still a hot challenge in research community Solution: We are trying to address this issue using directory searches like DMOZ or Google directory search
Digital Enterprise Research Institute Future Work Lots of enhancement can be considered See previous slide Developed as a hobby Later it attracted Ecospace consortium Tomorrow it will be demonstrated in the review meeting
Digital Enterprise Research Institute Have a nice day! Thank you! Try tools yourself: Questions?