SIGIR 2007 Proceedings

Doctoral Consortium

Fuzzy Temporal and Spatial Reasoning for Intelligent Information Retrieval
Steven Schockaer t


Ghent University Krijgslaan 281 - S9 9000 Gent, Belgium

ABSTRACT
Temporal and spatial information in text documents is often expressed in a qualitative way. Moreover, both are frequently affected by vagueness, calling for appropriate extensions of traditional frameworks for qualitative reasoning about time and space. Our research aims at defining such extensions based on fuzzy set theory, and applying the resulting frameworks to two important kinds of intelligent information retrieval, viz. temporal question answering and geographic information retrieval. Categories and Sub ject Descriptors: I.2 [Artificial Intelligence]: Knowledge Representation Formalisms and Methods; H.3 [Information Storage and Retrieval]: Information Search and Retrieval General Terms: Algorithms, Experimentation Keywords: Temporal Question Answering, Geographic Information Retrieval, Fuzzy Set Theory Recently, there has been growing interest in information retrieval tasks that require a thorough understanding of the meaning of a user request. Two particularly interesting examples are temporal question answering (QA) systems, which provide answers to temporally restricted questions like What major conflicts took place during the Cold War?, and geographic information retrieval (GIR) systems, which support geographically constrained queries like hotels in the old centre of Amsterdam. Such systems rely heavily on temporal and spatial information found in text documents, which is often affected by vagueness. For example, the time spans of many events (e.g. the Cold War, the Great Depression, the Dotcom Bubble, etc.) and the spatial extent of many regions (e.g. the Alps, Amsterdam's old centre, Western Europe, etc.) are characterized by inherently gradual boundaries, while metric information tends to be expressed in an approximate way (e.g. the hotel is located within walking distance of the Dam Square). In our work, we have defined a generalization of the Interval Algebra [1] and the Region Connection Calculus [2] to model vague temporal and vague spatial information respectively [5, 3], using fuzzy set theory as the underlying framework for representing vagueness. We have furthermore inResearch Assistant of the Research Foundation - Flanders. Steven.Schockaert@UGent.be
Copyright is held by the author/owner(s). SIGIR'07, July 23­27, 2007, Amsterdam, The Netherlands. ACM 978-1-59593-597-7/07/0007.

troduced sound and complete reasoning procedures for both generalizations [4, 6]. Our aim is to apply these reasoning procedures to improve the performance of temporal QA and GIR. For example, temporally restricted questions like Which battles were fought in Belgium between D-Day and the unconditional surrender of Germany? are usually tackled by extracting a list of battles that were fought in Belgium, determining when these battles took place, and comparing these dates with the dates of D-Day and the unconditional surrender of Germany. However, this strategy fails when no appropriate time span or date can be found for some of the events involved. We propose to use fuzzy temporal reasoning to deduce which answer candidates satisfy the temporal restriction in this case. To this end, we construct a large knowledge base offline, consisting of events, qualitative temporal relations between these events, and (possibly vague) time spans for as many events as possible. Similarly, it may not be possible to automatically extract the address or geographical coordinates of a particular businesses from the web. Fuzzy spatial reasoning can then be used to infer the relevance of these businesses to a geographically constrained query. Some of the most important issues we are currently facing are how to extract a sufficiently high number of reliable and relevant qualitative temporal and spatial relations from texts, how to obtain useful information about the approximate size of a geographic region (e.g. based on its semantic type), and how to evaluate the resulting systems.

1. REFERENCES
[1] J.F. Allen. Maintaining knowlegde ab out temp oral intervals. Communications of the ACM, 26(11):832­843, 1983. [2] D.A. Randell, Z. Cui and A.G. Cohn. A spatial logic based on regions and connection. In Proceedings of the 3rd International Conference on Know ledge Representation and Reasoning, pages 165­176, 1992. [3] S. Scho ckaert, C. Cornelis, M. De Co ck, and E. Kerre. Fuzzy spatial relations b etween vague regions. In Proceedings of the 3rd IEEE Conference on Intel ligent Systems, pages 221­226, 2006. [4] S. Scho ckaert, M. De Co ck, and E. Kerre. Qualitative temp oral reasoning ab out vague events. In Proceedings of the 20th International Joint Conference on Artificial Intel ligence, pages 569­574, 2007. [5] S. Scho ckaert, M. De Co ck, and E. Kerre. Fuzzifying Allen's temp oral interval relations. IEEE Transactions on Fuzzy Systems, to app ear. [6] S. Scho ckaert and M. De Co ck. Reasoning ab out Vague Top ological Information. Submitted.

921