Long running mailing lists and online fora are examples of rich but unstructured knowledge bases. This means that list subscribers have to do a lot of work to discover previous relevant discussions and conclusions.
Currently several ad hoc mechanisms exist for sifting out information from these archives, including the creation of FAQs (c.f. Dave Pawsons excellent XSLT FAQ) or summaries of "hot topics" (c.f. the XML-Deviant archives).
These mechanisms while useful are not satisfactory. In particular they rely on the abilities of one or two individuals to synthesise useful information from the community discussions. These individuals effectively act as editors of the community archives.
A more scalable mechanism would be one that could harness the community itself, or even a wider network of interested individuals, to edit and organise the knowledge base into useful categories.
This kind of approach lends itself to bottom-up (emergent) topic creation rather than a top-down approach. In fact the nature of discourse, and the evolving range of topics covered suggests that the latter approach is in fact unworkable.
This document sets out a proposal for a community topic building facility.
The system should:
- Operate independently of the list infrastructure (i.e. the archives, or mail traffic)
- Allow creation of new topics as simply as possible
- Allow arbitrary commentary and links to be added to a topic
- Allow the following relationships between topics to be made/removed:
- Allow any archived email to be associated with one or more topics. The actual mechanism should operate on URLs to allow any archive to be linked into the topic.
- Allow viewing of topics, displaying commentary, links to associated emails, etc.
- Allow topic-email links to be "promoted" from simple associations to a more refined commentary, e.g. so XML-Deviant style articles can be evolved from link associations. Basically topic editing to create articles.
- Provide a view of the topic structure, e.g. as topic maps/RDF/XFML for further processing
- Provide both machine-processable and human-reable views of the data.
- Provide APIs for above functionality allowing authoring tools, etc to be tied into the database
The topic structure can be built and annotated using a Wiki. Relationships between topics can be simple in-Wiki links. It may be necessary to extend the Wiki syntax to allow clearer delimiting of topic relationship links, and external links, etc.
A standalone Trackback server can be used to allow association between online documents and Wiki topic pages. Each topic page should have a trackback URL. This means it can be targetted from blog infrastructure.
A bookmarklet can be used to allow a user viewing an archived email to automatically ping the Trackback server and associate that email with the designated topic. The user can be prompted for a topic name (possibly including a pick-list) and commentary.
The bookmarklet may also be sensitive to selected text in the email, allowing quotes to be extracted from the page itself.
The Wiki page would include links to all emails/documents for which Trackback pings have been received. These may be displayed separately to the main page body (e.g. using a simple server side include) allowing Wiki commentary to be distinguished from commentary elsewhere. This means a community can be built around the Wiki and individual topic areas separate to the mailing list.
An optional feature would be to remove links to Trackback sourced document if these documents are also included in the Wiki page.
Another optional feature would be to allow notifications for page changes (both trackback links and inline commentary) to be sent to interested subscribers.
Another nice feature would be to include trackback data to be embedded in the mail archive, so the bookmarklet could just send pings, and the metadata (e.g. email author, subject) could be extracted from the archive itself.
- Trackback Standalone implementation
- Wiki for building RDF schemas
- Ridiculously Easy Group Forming