Cart
[Home] [By Thread] [By Date] [Recent Entries]
IF it is really a full-text indexing system, it scans and infers topics the same way a human scans and tags. It would require a rule base perhaps similar to a Schematron assertion engine. Past attempts made several passes over content to create a series of tagged documents that are successively refined. However, as in memory-based patterning systems, the more abstract the links, the more opinionated the system. Such systems can become very superstitious in exactly the same way people do. How was it phrased: "A schema is an opinion about a document..." http://www.answers.com/main/ntquery?method=4&dsid=2222&dekey=Memory-predicti on+framework&gwp=11&curtab=2222_1&linktext=memory-prediction%20framework len From: Robert Koberg [mailto:rob@k...] > Which is why I'd propose defining a full-text schema language, > so XML content can be described to a full-text search engine. It does sound very interesting. How would it work? What would it look like? I have tried doing this with XML Schema but gave up. I had tried to use annotations to give weight to different things, then I tried to make a type system. For me, it was just easier to write java to handle it. Now I write org.xml.sax.ext.DefaultHandler2's that suit my needs. I know, not very scalable or user friendly.
|
