Our research in the DAML program is to develop an ontology for representing temporal information. Temporal information is ubiquitous in real world situations. It includes dates of events (“January 3, 2002,” “next Wednesday”), durations of activities (“drive for twenty minutes,” “wait about an hour”), ordering between events (“wait for Fred then drive to Rochester”), and constraints between events (“don’t touch the button while the switch is on,” “only one flight can use the runway at a time”), among other things.
Our approach is to build the ontology bottom-up, by looking at how temporal information is described in natural language, and then abstract from that to the ontological categories. So far, we are concentrating on building a useful ontology starting from the Interval Temporal Logic or (Allen, 1983; Allen & Ferguson, 1997). In this logic, intervals of time are taken as primitive and relations are defined between them, such as “Before,” “Contains,” and “Disjoint.” The expressivity and computational properties of the logic have been explored extensively in the literature. This logic has proven useful in applications ranging from representing the content of natural language to planning courses of action in realistic domains. In many cases it is more natural to use the interval-based representation rather than a representation based on endpoints, because the interval relations can represent primitively situations that require disjunction to represent using endpoints.
Our technical goal for FY02 is to develop a practical DAML ontology based on the ITL concepts. Our initial plan is to reify the ITL relations into a set of categories that capture both ordering and metric information. For example, an instance of one of these might be “Airplane A arrived one hour before airplane B.” This includes the events and their ordering relationship (“A arrives” BEFORE “B arrives”) as well as a metric constraint on the distance between them (“one hour”). The other ITL relations can be similarly reified and extended. Our belief is that the power of the interval relations, and particularly their implicit use of disjunction, will be effective for representing complex temporal relationships within a DAML specification.
In terms of activities associated with the DAML Experiment, we believe that users (Subject Matter Experts) should be able to use our temporal ontology to capture and distribute information about the DAML Experiment scenario. This might include such things as courses of action, logistics requirements and schedules, observation and surveillance data, and so on. To help users, we will produce a “User’s Manual” for the ontology, describing its elements with examples of their use.
There are really three categories of metrics that are appropriate to consider for this work. The first, usability, is subjective, and amounts to whether people can use the ontology, how comfortable they are using it, and so on. This can be evaluated using questionaires and interviews. It can also be broken down and quantified more precisely by measuring coverage and reliability. Coverage is a measure of how much of the SME data can be represented using the ontology. Inter-annotator reliability is a measure of how likely different users of the ontology are to encode the same information using the same terms. For either of these, one would want multiple users encoding the same facts, to allow comparison.
The next logical step for temporal ontologies in DAML following our initial work is clearly to increase the expressivity of the language. One can already think of various forms of temporal information that might be difficult to represent using the “ordering + metric” relations we described previously. Disjunction is critical, particularly for representing uncertainty in the temporal aspect (e.g., “either A goes before B, or B waits for C”). Iteration is also very convenient (e.g., “repeat the signal up to five times,” “turn the screw until it’s tight”). There are probably others.
Next, there is the question of a deeper mathematical relationship between the metric and qualitative parts of the representation. There has been work on this problem in the temporal logic literature, but many questions are still unanswered.
Finally, one would want to develop reasoners that could perform inference using the elements of our temporal ontology. That is, from instances of certain relations, one would like to draw conclusions about the existence of other instances. Or one might want to check that a set of instances described a situation consistently. Some of this could probably be built on top of algorithms developed for the Interval Temporal Logic. The challenge would be to integrate the metric information in ways that do not make the problem to complex to compute.
Allen, J.F. (1983). Maintaining Knowledge about Temporal Intervals. Communications of the ACM, 26(11), pp. 832-843.
Allen, J.F., and G. Ferguson, (1997). Actions and events in interval temporal logic. Oliviero Stock, ed., Spatial and Temporal Reasoning, Kluwer Academic Publishers, pp. 205–245.