Considering Twine and Semantic Computation

Reading my feeds, I came across a review of Twine from an AI perspective titled: Why I Migrated Over to Twine. I had read about it before, and paid attention to the post on O'Reilly Radar, although I thought I would have a deeper look and think about it.

The product appears to be a wiki like knowledge management system requiring the user to load in their content explicitly. The expectation appears to be that future content creation (documents, media, etc) occur from within or in conjunction with the application. It also seems to provide all the social network features one would expect for relating a user to additional content, people, and interest groups.

Twine is a semantic application, in that they claim to use semantic understanding to learn about the users interests. The product of the learning is exploited through recommendations, connections, and ultimately productivity and or user experience. The system processes text for resources like peoples, places, things, and uses classical (W3C) semantic technologies such as RDF (subject, predicate, object) for resource description, and OWL to assign resources properties. Interestingly they seemingly provide REST-like accessibility of resources, and mined Wikipedia to seed their knowledge base. One thing that I am interested in are the technical ways in which the semantic data from extracted resources is maintained (does a RDBMS suffice?).

The site has seen a lot of press, among which was it's review and relative comparison with the state of semantic web applications late last year on RWW.

To me, the slow take-off of the semantic web feels a lot like the slow take-off of DVD. It was hyped and talked about for what seemed like years, creeping into shops and homes, until before you knew it, DVD was all any body wanted or could acquire. The semantic web feels as though it is doing the same thing, only it is strange because this is not the way information technology operates. Attrition of technology and standards is ruthless, and one would think that slow adoption would result in a similar state of affairs as that of SOAP (adoption constrained to enterprise in favor of lean alternatives).

Informed users are screaming for solutions to the overload of information, and informed computation (based on the meaning of information) is considered to be a direct solution. It is hard to think out effective uses for semantic computation beyond recommendation and discovery, at least off the cuff. I think the Twine application appears interesting, for example I would love to have a play (when they reach me on the list of beta users), although the lock in of the application for personal knowledge management would be a concern for me. It begs the question of whether centralisation of user content (as was also the case with lifestreaming) is an effective path to exploiting semantic computation. I think this would only be the case if the semantic-driven features offered outweigh the required churn.