Discovery of T-Templates and Their Real-Time Interpretation Using Theme
The temporal structure of every-day human behavior and interactions is certainly a complex affaire rich in repeated patterns or translation symmetry. This paper concerns a view of the structure of real-time streams of behavior as repeated, temporal patterns of a particular kind called t-patterns. An instance of a pattern of this kind consists of a particular and possibly quite small set of primitives of behavioral significance (verbal, nonverbal and/or environmental) occurring significantly more often than chance expectation in a particular order and/or concurrently with characteristic intervals between them. The analogies thus exist with speech and writing where only a few letters or phonemes are combined to create hundreds of thousands of different words and common word combinations. While remaining statistically significant, the time structure of t-patterns is also flexible and thus accommodates that of, for example, words, phrases, melodies and musical themes, which may be performed with considerable variation between repetitions in speed and internal intervals.
T-patterns are thus hierarchical patterns of patterns etc. and as phrases, their interpretation and effects (meaning, function) depend on the particular words involved, the temporal aspects of their production (performance) and the general context in which they occur. Some t-patterns occur cyclically and this better known aspect is now also automatically detected by the software THEME, which has been specially developed for t-pattern detection. A typical characteristic that has caused much difficulty regarding the detection of behavioral “sequences” is that routines, ceremonies and verbal “t-frames” (such as if.. then.. else) is that other behavior may occur in various numbers and ways between the components of different instances of the same pattern. Profiling individuals, interactions and/or groups can be based on the existence of particular t-patterns and/or the absolute or relative frequencies of patterns.
The THEME software also detects various other phenomena derived from the so called critical interval relationship and the t-pattern type such as t-bursts, t-cycles, t-markers, t-paths, t-associates, t-frames, and t-packets. This is the primary task of Theme, but a considerable part of the software helps with the analysis and use of the detected patterns, which is done both through visual and statistical means.
The software can thus automatically analyze a large number of datasets in a single run and automatically build a data base of detected patterns that can then be consulted in various ways. Highly significant effects of independent (experimental) variables on the frequency and complexity of detected t-patterns have often been found in studies where no significant effects were detected using the same initial data and standard statistical methods alone.
Real-time use of THEME for the interpretation of ongoing behavior seems feasible given some further development. Theme thus already automatically creates t-pattern templates (t-templates) on the basis of a detected pattern base — which can be updated off-line from time to time. Template matching being much faster than the preceding pattern discovery and template construction, t-templates could be matched real-time against incoming data. Higher speed could be obtained through parallel processing. THEME is currently developed in Delphi 2005 Professional and large parts have already been transferred to Linux using Kylix (for use in Bioinformatics) partly in preparation for a parallel processing version.
KeywordsTemplate Match Ambient Intelligence Standard Statistical Method Hide Structure Markov Chain Analysis
Unable to display preview. Download preview PDF.
- Anolli, L., Duncan Jr., S., Magnusson, M. S., and Riva, G. (Eds.) (2005). The Hidden Structure of Interaction: From Neurons to Culture Patterns. Amsterdam: IOSPRESS (View/Download PDF version).Google Scholar
- Bakeman, R. and Gotman, J. M. (1997). Observing interaction: An introduction to sequential analysis. Cambridge: Cambridge University Press.Google Scholar
- Eibl-Eibesfeldt, I. (1970). Ethology: The Biology of Behavior”. New York: Holt, Rinehart & Winston.Google Scholar
- Magnusson, M. S. (2000). Discovering Hidden Time Patterns in Behavior: T-Patterns and their Detection. Behavior Research Methods, Instruments and Computers, 32(1): pp. 93–110. (View/Download PDF version).Google Scholar
- Magnusson, M. S. (2003). Analyzing complex real-time streams of behavior: repeated patterns in behavior and DNA. L’éthologie appliquée aujourd’hui. (C. Baudoin, ed), Volume 3–Ethologie humaine. Levallois-Perret, France: Editions ED. ISBN 2–7237-0025–9.Google Scholar
- Magnusson, M. S. (2004). Repeated Patterns in Behavior and Other Biological Phenomena. In Evolution of Communication Systems: A Comparative Approach (Vienna Series in Theoretical Biology). D. Kimbrough Oller (Editor), Ulrike Griebel (Editor). London: The MIT Press. ISBN: 0262151111 (View/Download PDF version).Google Scholar
- Magnusson, M. S. (2005). Understanding Social Interaction: Discovering Hidden Structure with Model and Algorithms. In Anolli, L., S. Duncan Jr., M.S. Magnusson and G. Riva (Eds.) (2005). The Hidden Structure of Interaction: From Neurons to Culture Patterns. Amsterdam: IOSPRESS (View/Download PDF version).Google Scholar
- Magnusson, M. S. (2006). Structure and Communication in Interactions. In G. Riva, M. T. Anguera, B. K. Wiederhold and F. Mantovani (Eds.) From Communication to Presence: Cognition, Emotions and Culture towards the Ultimate Communicative Experience. Festschrift in honor of Luigi Anolli. IOS Press, Amsterdam, 2006. (View/Download PDF version.).Google Scholar