KIWI after one year from IWIS group perspective

Peter Dolog's picture

A bit late, a week after annual meeting, but still I have decided to write a post as my view on KIWI project after one year. I think the KIWI project had a very successful first year. There is a release of the KIWI platform already now available and there is a large ongoing activity to further develop it. It is interesting how the thoughts in the project developed. We started from the semantic wiki platforms it was understood that time. Pages, typed links, and RDF annotations have been the main entities used in querying and reasoning. Soon we realized that perhaps tagging would be easier for a user of any kind to maintain and edit. Therefore, KIWI prerelease of course has tagging in the release . The tagging also changed our view on personalization a bit. Before we wanted to experiment with pure logical reasoning over user explicit metadata and content explicit metadata. Over the year of experimentation we have understood that the tagging as carrier of context and knowledge about the purpose of a content and user is rather of more evolving nature than traditional encoding of metadata with existing ontology. The reasoning in such a context is then rather based on multiple factors calculated from several views on the data represented rather in a traditional algorithmic way and automated reasoning perhaps adds to the results of the pre-computation. IWIS team in particular developed special widgets for three kinds of content items recommendations based on different factors. The evolving nature of the KIWI content and in general any enterprise content reflected in the information extraction work as well. People focusing this area in the KIWI project has provided an on the fly annotation component used to derive annotations from text when user is writing instead of focusing on batch jobs. This has two advantages: it does not need a large training set and user can immediately validate the results. Last but not least, the use cases developed into two very interesting focus areas: enterprise community support and interplay of social evolving content and structured data about software work products. But we did not have only success. Enterprise mesh ups with semantic wikis with user centered programming was an idea I particularly like and I think it has a future. We have not progressed in this direction very much and it seems that we will have to leave this to other projects to spare resources to develop those interesting things which are already in place in some form . So, happy birthday KIWI and I am looking forward to the forthcomming two years.