|
|
| KMi | Scholarly Ontologies Homepage |
 |
ClaiMaker
Executive Summary
A tool
for the Scholarly Ontologies Project made at KMi as part of the EPSRC
Distributed Information Management Programme (GR/N35885/01: Duration: 2001-2004)
|
 |
Problem: Technology is changing the way we publish documents. Digital libraries and pre-print archives potentially give unprecedented access to texts and data collections. These are well served by text search engines, but do not support the interpretation of documents. The Web has illustrated how non-linear formatting can enhance information. In this new world, scientists and academics, released from the constraints of the printed page, can contemplate new ways to disseminate their results.
Approach: The Scholarly Ontologies project is developing and assessing a tool, ClaiMaker, for modelling documents, that combines knowledge engineering and visualisation. A community of users (who may be the original authors or readers) collaboratively builds a model of the core ideas in their literature and the links between them. We call two ideas joined by a link claims. Claims are connected to online versions of the documents, and so ClaiMaker can be used as an index to a digital library, as well as a tool for the communication and interpretation of research.
At the core of the Scholarly Ontologies approach is a discourse ontology, a structured set of suitable links to be used by authors when making claims and positioning them with respect to earlier work. The discourse ontology has several benefits. First, in the context of creation, it provides a common language that the community can use to construct their models. Second, the ontology provides features that are being used to design innovative search, interpretation and visualisation tools to allow users to navigate large and growing models.
Vision: Initial work is focused on scholarly publishing, but we envisage that the approach would work well for corporate knowledge analysis, or in education, where students might model a course reading list, or a research literature review. Our strategy is that a group in a discipline catches the vision of a new, 'net-centric' way of publishing and analysing research claims. We're bootstrapping a new publishing medium here, so it's likely to be 'early adopters/researchers' at this stage. Having modelled part of their literature in the system, and demonstrated to their own satisfaction its potential as a way to publish and analyse the connections between ideas, this can serve as a relevant exemplar for communicating the idea to their wider community.
Work in brief: In our prototype, ClaiMaker, we are putting these principles into practice. The claim creation interface introduces users to modelling with a familiar form-filling approach.
Figure 1: Claim creation interface in ClaiMaker, showing how a researcher can build a set of claims. Key: [1] a claim that has already been constructed, ready to submit, [2] a concept to link from, [3] a link selected from the discourse ontology
The Scholarly Ontologies project is also exploring ways of presenting the models in our discovery tools. This listing approach is taken from the Neighbourhood search:
Figure 2: Results listing for a Neighbourhood search in ClaiMaker. Key: [1] icons bring up further information about the concept, [2] link direction is indicated by the arrows, and green, gold and red arrows indicate that a link has positive, neutral or negative implications.
Future work will exploit the basic graph structure of the claims database to produce visualisations of the data. This diagram is a preview of one possible look & feel for the Lineage search (currently under development).
Figure 3: A possible visualization of the results of a lineage search on a concept (top node), showing two branches. (Graph visualization courtesy the Ceryle Project by Murray Altheim, Knowledge Media Institute, Open University. Ceryle includes an enhancement of the TouchGraph graph visualization toolkit by Alex Shapiro. TouchGraph delivers interactive, self-organizing maps over the web via a Java applet.)
