At the Linked Open Data in Libraries, Archives and Museums conf [LODLAM], Jonathan Rees casually offered what I thought was useful a distinction. (Also note that I am certainly getting this a little wrong, and could possibly be getting it entirely wrong.)

Background: RDF is the basic format of data in the Semantic Web and LOD; it consists of statements of the form “A is in some relation to B.”

My paraphrase: Before LOD, we were trying to build knowledge representations of the various realms of the world. Therefore, it was important that the RDF triples expressed were true statements about the world. In LOD, triples are taken as a way of expressing data; take your internal data, make it accessible as RDF, and let it go into the wild…or, more exactly, into the commons. You’re not trying to represent the world; you’re just trying to represent your data so that it can be reused. It’s a subtle but big difference.

I also like John Wilbanks‘ provocative tweet-length explanation of LOD: “Linked open data is duct tape that some people mistake for infrastructure. Duct tape is awesome.”

Finally, it’s pretty awesome to be at a techie conference where about half the participants are women.

3 Comments »

Let me disambiguate that title: Reuters is offering a Web service, called Calais, that will parse text and return it in a form (RDF) that can be utilized by Semantic Web applications. It uses natural language processing (from ClearForest) to find structures of meaning such as places, jobs, facts, events, etc. It apparently has its own metadata schema, but it allows users to extend it. It’s an open API, and Reuters is being quite generous in how much they’ll let you submit during this beta period. It’s English only for now, although they plan to support other languages, opening the exciting prospect of being able to find items of interest in languages you don’t understand via a unified metadata framework.

I’m going by the site’s FAQ. I haven’t tried it and can’t tell how well it works, how accurate it is, how comprehensive or detailed its metadata are, and how much post-processing cleanup uses will want to provide (which of course depends on the application). There are some points I just don’t understand, such as the claim “Calais carries your own metadata anywhere in the content universe.” But if it works within some reasonable definition of “works,” and if it gets widely adopted, Calais could make a lot more information a lot easier to find, and to process for further meaning. [Tags: semantic_web semweb everything_is_miscellaneous reuters calais nlp ]

3 Comments »