Determining relevance of imprecise temporal intervals for cultural heritage information retrieval

Abstract

Time is an essential concept in cultural heritage applications. Instances of temporal concepts such as time intervals are used for the annotation of cultural objects and also for querying datasets containing information about these objects. Hence it is important to match query and annotation intervals by examining their similarity or closeness. One of the problems is that in many cases time intervals are imprecise. For example, the boundaries of the “Pre-Roman age” and the “Roman age” are inherently imprecise and it may be difficult to distinguish them with clear-cut intervals. In this paper we apply the fuzzy set theory to model imprecise time intervals in order to determine relevance of the relationship between two time intervals. We present a method for matching query and annotation intervals based on their weighted mutual overlapping and closeness. We present (1) methods for calculating these weights to produce a combined measure and (2) results of comparing the combined measure with human evaluators as a case study. The case study takes into consideration archaeological temporal information, which is in most cases inherently fuzzy, and therefore offers a particularly complex and challenging scenario. The results show that our new combined measure that utilizes different weighted measures together in rankings, performs the best in terms of precision and recall. It should be used when ranking annotation intervals according to a given query interval in cultural heritage information retrieval. Our approach intends to be generalizable: overlapping and closeness may be calculated between any two fuzzy temporal intervals. The presented procedure of using user evaluation results as a basis for assigning weights for overlapping and closeness could potentially be used to reveal weights in other domains and purposes as well.

Introduction

Time is one of the central concepts in ontologies representing the world, and hence should also be centric in annotations and queries of the Semantic Web. Time is especially important for managing historical collections, for example in visualizing them on a timeline (Hyvönen et al., 2009, Schreiber et al., 2006).

However, representing time in Semantic Web ontologies is not straightforward because the question of when a certain time was or will be is often uncertain, subjective or vague (Nagypál and Motik, 2003). For example, it may not be known when exactly a given archaeological artifact was manufactured (uncertainty), when “The Middle Ages” was according to opinions of different historians (subjectivity), or when the spring starts (vagueness, imprecision).

In addition, transitions between different phases, such as historical periods, are usually complex processes which are not identifiable by clear cut dates, even if conventional calendric markers are mostly used in order to simplify historical sequences. All these elements are at the basis of imprecise temporal representations especially in the cultural heritage, historical and archaeological contexts.

Nevertheless, representations of time are needed for representing and matching annotations and queries. The definition of temporal intervals and their relations are crucial in the context of archaeological chronologies. Checking whether two time intervals have something in common allows for answering queries like “find all artifacts manufactured around the middle of the I century B.C.”. At the same time, the often inherently fuzzy nature of historical and archaeological temporal information makes this scenario particularly complex and challenging.

In this paper we adapt the idea (Nagypál and Motik, 2003, Visser, 2004) of using fuzzy sets (Zadeh, 1965) for the representation of temporal annotations and queries. The structure of the paper is as follows. In Section 2 we present a new method for matching queries and annotations using their weighted mutual overlapping and closeness. In Section 3 we present a case study where the method has been implemented and provide results of its evaluation with human test subjects. The results in Section 5 suggest that a measure that combines different single, weighted measures together performs best in terms of precision and recall, and should hence be used when ranking annotation intervals with respect to a given query interval. In Section 6 we provide a discussion about the results, and Section 7 provides discussion about the related work. Finally, Section 8 concludes the paper.