A Pattern for Periodic Intervals

Tracking #: 897-2108

Maria Poveda
Mari Carmen Suárez-Figueroa
Asunción Gómez-Pérez

Responsible editor: 
Krzysztof Janowicz

Submission type: 
Ontology Description
Non-convex intervals allow identifying periodic intervals with gaps between them (e.g., “every Wednesday”), while convex intervals are those that are not composed of “separate pieces” (e.g., “from 1st April 2014 to 30th April 2014”). Non-convex intervals consist intuitively of some convex subintervals with convex gaps in between. In this paper, we describe an ontology design pattern to represent periodic intervals, a specific case of non-convex intervals in which the period between subintervals and the duration of such subintervals are constant.
Full PDF Version: 


Solicited Reviews:
Click to Expand/Collapse
Review #1
Anonymous submitted on 25/Dec/2014
Review Comment:

This paper describes an extension to the OWL-Time ontology for Periodic Intervals, defined as intervals containing sub-intervals of the same length at a recurring period length. On its face this seems to be a clear gap in OWL-Time, and worth filling. A number of issues identified by reviewers of the first draft have been resolved. Two fundamental concerns remain however.

First is the fact that months and years have irregular lengths, which differentiates them fundamentally from seconds, minutes, days, and weeks. No two months are necessarily of the same duration. Years defined astronomically as earth orbits can be said to be essentially of the same length, but years in the calendar days of ISO-8601 vary in length every fourth year. For these reasons, the same kinds of computation could not be performed on intervals involving years and months as on the other four Duration Descriptions. Does the pattern/extension remain useful in this case?

This leads to a second concern, concerning potential operations on these temporal objects more generally. The list of two competency questions in Section 3 is too short. Only one refers to a logical operation, and in that case, the ontology as described would be unable to answer it (there is no mention of days of the week).

For these reasons, I think the premise of the paper—that the pattern for Period Intervals as modeled will serve a sizable number of cases—is flawed, and needs to be reconsidered.

Review #2
By Simon Cox submitted on 03/Jan/2015
Minor Revision
Review Comment:

The manuscript has been significantly revised and expanded, and the minor issues that I raised have been resolved.

Nevertheless, I still have concerns related to the two major issues raised in my original review.

1. The primary concern (that the term 'period' was defined in a way that was inconsistent with more general usage) appears to have been rectified, though it requires a fairly careful reading of the paper to discern this (Figure 3 and 6). The last part of 8th sentence in the 4th paragraph should be clarified from "(b) the periods between them are also equal." to "(b) the periods between the start of each component interval have equal duration."

The conceptual model and its implementation in the ontology are both shown graphically in Figure 3. I suggest that Figure 1 be revised to show the conceptual elements (i.e. the terms denoting the green ellipses in Figure 3). In this way the conceptual model is established first, independent of its implementation as an extension to OWL-Time.

Finally, there is a spelling error in Figure 3: "l3 beggining" should be "l3 beginning".

2. I remain concerned about the use of the term 'Pattern' in the title. A pattern is a solution that may be applied to multiple problems. It is not clear how that applies to this ontology, which extends and re-uses OWL-Time to solve an additional problem in a compact manner. If the same solution was also applied to (say) a periodic spatial phenomenon, then it might be called a 'pattern', but as it stands there is no demonstration of how it could solve related problems. The title shoudl be changed to something like "An extension of OWL-Time for Periodic Intervals".

Review #3
By Carsten Keßler submitted on 06/Jan/2015
Major Revision
Review Comment:

The paper has improved singnificantly in terms of presentation and illustration.

Having said that, I am under the impression that the pattern cannot be used to represent periodic intervals that do not have a specific beginning/end. For example, if I wanted to say that the German public holidays for Christmas start at 1PM on December 24th, and end at midnight on the 26th *every year*, this is not possible without specifying when the overall periodic interval starts, because then the start of the interval/period would be undefined. In that case, I could only say that there is a ~2.5 day long interval of Christmas holidays in a 1 year long period, but I could not say when the interval/period starts. Even the simple examples “every Monday” and “A pharmacy whose opening hours are from 9:00 to 21:00 every day“ mentioned in section 3 cannot be handled without assigning a start and end for the entire periodic interval.

This seems to be a conscious design choice, as the cardinality of the properties owl-time:hasBeginning and owl-time:hasEnd have been set to 1. However, it does seem to be a shortcoming to me, and the examples mentioned above that are explicitly given in the paper evoke the impression that the pattern can handle such unbounded periodic intervals. I may be missing something here, but if the pattern can indeed handle these intervals, a simple example should be included (e.g., for “every Monday”) in addition to the GoT example. In any case, I think that this is a serious problem that needs to be addressed before the paper can be published.

Review #4
By Tomi Kauppinen submitted on 14/Jan/2015
Review Comment:

This new version of the paper is significantly improved and targets all the comments I had.