ReTAX+ fact-file

What's the Problem?

Taxonomies are an important form in which some sciences (eg botany) organize aspects of their knowledge. Creating a consistent taxonomy is an important task for botanic taxonomists. Much work was done in the 19^th century in creating these taxonomies initially, but now they have to be kept consistent both when new species are discovered and when the field decides to introduce new descriptor(s) (say available from a novel kind of instrument). This is a complex task which needs a great deal of computer support, [Middleton & Wilcox, 1990]. Further we believe that many of the techniques evolved here could also be used to help Knowledge Engineers enhance / refine Ontologies.

Towards a Solution

ReTAX+ is a recent Java implementation of ReTax, [Alberdi & Sleeman, 1997] which was originally implemented in LISP. Both implementations incorporate rules which determine whether a taxonomy is well formed:

• each child node must be more specialized that its parent
• each of a node’s siblings must be unique

The functionality provided by RETAX

Let's refer to a new object/node as N, the existing hierarchy/tree as T, and the potential parent node as P. Then possible operations are:

• Is T well formed? (If not report nodes which violate the rules)
• Is N already in T?
• Assuming T is well-formed, to which parent node, P, can N be attached without causing T to be rearranged or N modified? (Answer could be none)
• What changes have to be made to N to make it a 'legal' child of node P?
• What changes have to be made to T so that N can be a child of P?
• Combinations of the last 2 operations

We would argue that the above functions enable the taxonomist to do a series of WHAT IF experiments to see whether the new entity can be fitted into the tree, T, without change or what changes need to be made both to the entity and the tree to accommodate it satisfactorily.

We have a 'vehicles' data-set and associated taxonomy (available in Powerpoint)

Suppose each of the objects are described by the following fields / descriptors:

• name / label (a string)
• number of wheels (integer in range 1-8)
• size: ordered-set (low, medium, large, or high)
• motor (yes or no)
• engine-power (integer in range 0-20)

So that if we assume 'reasonable definitions' for these objects which can be found in PPT, and if we attempt to introduce the following new entity:

• name: mono-cycle;
• number of wheels: 1
• size: low
• motor: no
• engine-power: 0

In the guided tour below we show the effects of each of the operations available in RETAX+ and listed above. So we demonstrate that the taxonomy provided is in fact well-formed, that the new entity (the mono-cycle) is not already in the taxonomy, T etc.

If we investigate the operation: What changes have to be made to T so that N can be a child of P? Where we specify that the selected node is bicycle then we see that RETAX+ will make changes to the node bicycle and then cycle.

ReTAX+: Future Plans

• Use with other experts to help them formulate / refine taxonomies (eg microbiology)
• Have Experts explain why they select certain options to accommodate a new entity and reject others (this is using RETAX+ as a sophisticated KA system)
• Use RETAX+, or a variant, to formulate / refine ontologies (eg medical terminologies)

Take a Guided Tour

A movie of ReTAX in action is available (TSCC codec required).

Further Reading

D.J. Middleton and C. Wilcox (1990) A critical examination of the status of Pernettya as a genus distinct from Gaultheria. Edinburgh Journal of Botany, 47, p291-301.

E. Alberdi and D. Sleeman (1997). ReTAX: a step in the automation of taxonomic revision. Artificial Intelligence, p 257-279.

E. Alberdi, D. Sleeman and M. Korpi. (2000). Accommodating Surprise and Taxonomic Tasks: The Role of Expertise. Cognitive Science, 24(1), pp53-91.

Semantic representation

View in the AKT Triplestore Browser or as RDF.

Also available in DOAP RDF (Description Of A Project)