It is well accepted that learning does not end with the completion of formal education. Much of the learning that occurs throughout life is supported not by textbooks or formal training programmes, but through community membership. Learning about a topic becomes synonymous with learning to be a member of a community of people who are experts on that topic. These communities range from workplace groupings that share ideas and stories, to hobbyists who form communities of shared interest. Increasingly, these communities are learning and sharing ideas at least partially online. These (wholly or partially) online communities create and share messages and documents that could form a useful learning resource for new and existing members. However, finding appropriate content from a large and growing resource, for example by employing text search engines, can often be inefficient or even fruitless.
Currently, a number of knowledge technologies are available aimed at alleviating this problem. Many involve annotating the documents and other artefacts of interest according to some type of formal representation or ontology of the knowledge and interests of the community. These approaches run into difficulty if the turnover of documents necessitates a level of maintenance beyond the available resources of the community. Maintaining a knowledge base involves effort and therefore cost. For the effort to be worthwhile, the cost has to be paid back, ideally with interest, as workers benefit from access to useful resources. In corporate enterprises, even when the knowledge contained in documents is valuable, if the topic of interest is fast moving then individual documents have a limited life span in which to pay back the investment. This is even more true in non-corporate communities where the required effort is supplied on a charitable basis, which is always a dangerous ingredient for groupware.
We aim to tackle the problem of fast moving information by leveraging the characteristics of the community that use and generate the resource. Our motivation for adopting this approach is that communities have particular characteristics, such as communicative genres, which are revealed in their way of doing things. We hypothesize that if this personality of the community could be understood it could be utilized in the development of low maintenance tools that could provide high accuracy for the search and retrieval of important community artefacts. More specifically, we hypothesize that the perspective of the community (i.e. what the community sees as important) will be reflected in the implicit structures found in the artefacts they create and share. Our approach utilizes and adapts a number of existing tools and methodologies including heuristic analysis, knowledge modelling and latent semantic indexing.
T.D. Collins, P. Mulholland and S.N.K. Watt. (2001) Using Genre to Support Active Participation in Learning Communities, In The Proceedings of the European Conference on Computer Supported Collaborative Learning (Euro-CSCL'2001), Maastricht, Holland. Available as KMi technical report KMi-TR-97
Category based search tool - A category based search tool for use with KMi Planet an online newsletter used in our lab. Based on the 'inverted pyramid' approach to writing news, stories are classified according to the vocabulary used in the story's first sentence.
Similarity based browsing tool - A browsing tool for the local history pages produced during the course of the Millennium CLUTCH Club. A web page is displayed along with links to similar pages, where similarity is measured according to a term frequency by inverse document frequency (TFIDF) metric.
May 2000 - May 2002
The kind of work carried out in the ELC project is now being continued in the CIPHER project. CIPHER is an acronym and stands for enabling Communities of Interest to Promote Cultural Heritage of European Regions. Further details on the CIPHER project can be found on the OU's CIPHER project page, or the main project website.