On Wed, Aug 08, 2001 at 10:45:43AM +0800, Stas Bekman wrote: > On Tue, 7 Aug 2001, Jim Smith wrote: > > > On Tue, Aug 07, 2001 at 10:16:26PM +0800, Stas Bekman wrote: > > > > Just some pseudo-random ideation boiling down to "let's use mod_perl > > > > to buils a knowledge base" both to demonstrate it's power and to serve > > > > the community. > > > > > > I like the idea!!! > > > > If we can come up with a proposal for a generic knowledge base product that > > would be useful in an IT environment, I can probably devote some of my work > > to it -- this is something I've been wanting to put together at work for > > customers and our help desk people but haven't had time yet to get it all > > together. I'll have more time in September after the students return. > > go ahead and be the first to propose Jim :) Ok... here are some of my initial thoughts. We need to be able to enter arbitrary documents, so I suggest DocBook as the standard format. This handles articles, books, reference pages, etc., in a well-defined manner. It also allows us to transform to other formats without much of a problem. We can even consider AxKit for at least part of the web interface. We would want to have different sections for documents that are not related. For example, we (here at TAMU) could use a section on Unix, another on Mac, and yet another on Windows systems. Or we could divide it up by services. The different sections would not expect overlap in keyword -> content mapping, so we could have an AI::Catagorize object for each section that could provide a default set of keywords. As we entered documents, those objects could learn which keywords were appropriate. We would want to have documents in multiple catagories. This might require the person entering the document enter multiple sets of keywords, one per section. We would need to index on keyword so people could quickly find the documents. Perhaps even a catelogue-style interface for browsing that would be based on keyword categories. This would require some work. (Broad categories are indicated by the presence of certain keywords, or by a weighted average so a document having all but a couple of the appropriate keywords won't be dropped from that category.) Documents shouldn't have to be entered via the web interface, though they could be. We could provide a set of web-page templates for each of the DocBook formats (well, the small ones anyway - don't want to write a book via a web interface). Might want to even integrate with WebDAV and a repository. We probably want to set up a SourceForge project if this is a go. Any ideas on names? Now to see if I can get my boss to support me spending time on such a project :) --jim
