Creating and Using Organizational Semantic Webs in Large Networked Organizations
Sam Chapman of k-now.co.uk and Sheffield University presented a discussion of how to get large organizations to share knowledge. He observes that many portals are abandoned in favor of ad-hoc methods like emailing data around or sharing directories. One big reason is that the portal is too rigid, forcing everyone to follow same schema. Need something more flexible, that lets individuals pick their own knowledge structures, but supports the imperfect linking of different structures in the organization. Hmm, what could we use. How about the semantic web? They have a framework in several parts:
- K-forms: visually author forms for data capture, and by doing so, you implicitly define an ontology of that data. Suggestions as you build, to use piece of other forms that appear to match, which encourages ontologic “convergence”. Each form defines an object type; fields are properties or relations. You can have subforms, which define other types held in relation to the starting type. Users fill in the form, and the data gets shoved into a spreadsheet.
- Legacy data handled by machine learning approaches to extract information from forms and from text.
- K-search offers ontology-based search, mixed with keyword search.
They did an evaluation with 6 users. System was easty to use. On average, 60% of available concepts (ontology elements) were reused—both people’s own concepts, and those original created by others. So they’ve been successful in encouraging ontological convergence.
The framework was used in weknowit, which captures information on responses to emergencies. Also tried at Rolls Royce, very popular there.
We’re starting to see lots of frameworks like this (create an ontology implicitly by building the forms you want to use for entering data). App2you is another example. But I haven’t seen this approach in the PIM space, where I think it could be very effective.