A Paradigmatic Perspective
4. Concluding Discussion
This cognitive approach to IR theory is supported by empirical evidence from both the infor-mation seeking research as well as from studies made in the operational online environ-ments. In addition, it does not contradict the IR developments achieved so far associated with the algorithmic research paradigm. It attempts to incorporate them all in a global frame-work.
Polyrepresentation of the user's knowledge Space signifies to represent not only the current information need, but in addition, and more importantly, to reach into the underlying problem space, actual work task or interest, and the dominating work domain(s). These elements are associated with the information need formation by following a principle of causality, that is, the intentionality underlying the fact of having such a need at all. Information needs internal to the user are basically regarded vague and variable, but may also exist in stable and well-defined forms, according to the nature of the current State of the user in the Situation. The representations of all elements in terms of request, problem and work task formulations may point to the actual condition of the intrinsic information need.
For relevance assessment the theory suggests and allows for relative, but statistical reliable assessments. Relative but unreliable assessments are often the case in retrieval
experi-ments, since the recall ratio, pre-defined per simulated request, is established in an arbitrary and uncertain way. In a statistical sense all laboratory experiments based on one person's initiary base-line relevance assessments are highly unreliable and insufficient since a sec-ond assessor would without doubt judge differently. In addition, human partial relevance as-sessments of documents is made available and necessary following the cognitive approach, since the semantic entities or passages, not the entire documents, often are the main objec-tive of retrieval, assessment and use. Hence, the theory also encourages to include differen-tiated relevance, implying that 'topicality' relevance is supplied with work task and problem relevance, or rather, forms of 'situational relevance' (Schamber et al 90). Several passages may be topically relevant, but only partially relevant or irrelevant for the actual work task and problem - and vice versa.
It is interesting that the partial or decimal-scaled relevance assessment has always been allowed the algorithmic retrieval engines, e.g. the vector Space and probabilistic modeis in their ranked Output. In the same modeis human assessments are supposed to be binary, as if humans are less distinctive than machines. This unrealistic (reductionistic and linear) sce-nario can be experimentally overcome by the cognitive framework (Borlund and Ingwersen 97).
Another intriguing point is the relatively high retrieval Performance produced by the two algo-rithmic modeis. Although quite different in their mathematical treatment of terms from infor-mation objects and in requests, both modeis explicitly break down any semantic coherence in a given object or request. The more rieh the request the better Performance of the modeis.
The seemingly unrealistic assumption behind the modeis is that words in texts are inde-pendent of one another. A second peculiar assumption for the probabilistic model is that rele-vance judgements of several information objects also are independent. Single terms from individual documents or passages, that are judged relevant, may thus form the ensuing query Version. Despite these limitations from a theoretical cognitive point of view these algo-rithmic modeis are indeed covered and explainable by the theory: The term independance can be seen as producing heavy inconsistencies because Single terms per se carry a very large number of semantic potentialities. This semantic openness is their asset operationally.
By combining a significant number of Single words the probability increases to retrieve ob-jects of füll text documents or passages in which a large portion of the input words actually is present. Then, the shorterthe document or passage containing any conceivable word com-bination, the higher the probability of semantic coherence (or meaning) in the passage re-trieved. This is due to the fact that text passages are generated by intentand imply meaning.
High topical relevance is thus achieved in a linguistic bottom-up approach. So, the term inde-pendence assumption is linguistically (and cognitively) sound. The condition is that the number of lexical terms that is applied is high.
The relevance assessment independence assumption associated with the probabilistic model is unrealistic in the case of assessments of lists of titles, since these are directly com-parable by the searcher. The quality of the assessments may in addition very well be poor.
However, in the case of longer passages or füll text documents assessed in total one by one, the assumption eoineides with the well-known phenomenon of cognitive information over-load: it becomes difficult or impossible to remember previous assessment results and their underlying rationale. The assumption consequently tend to be realistic (and true).
The current experimental situations can be enhanced by introducing real-life and non-simu-lated information need situations in order to prove their realistic reliability. The number of experimental variables will, as a consequence, increase further compared to today's re-search settings in IR.
References Beaulieu 96:
M. Beaulieu, S. Robertson and E. Rasmussen. Evaluating Interactive Systems in TREC.
Journal of American Society for Information Science, 47: 85-94,1996.
Belkin 90:
N.J. Belkin. The Cognitive Viewpoint in Information Science. Journal of Information Science, 16(1): 11-16,1990.
Belkin and Croft 87:
N.J. Belkin and W.B. Croft. Retrieval Techniques. Annual Review of Information Science and Technology, 22:109-145,1987.
Belkin et al. 82:
N.J. Belkin, R. Oddy and H. Brooks. ASK for Information Retrieval. Journal of Documen-tation, 38: 61-71 & 145-164, 1982.
Belkin et al. 87:
N.J. Belkin, H. Brooks and P.J. Daniels. Knowledge Elicitation Using Discourse Analysis.
International Journal of Man-Machine Studies, 27: 127-144,1987.
Belkin et al. 93:
N.J. Belkin, C. Cool, W.B. Croft and J.P. Callan. The Effect of Multiple Query Repre-sentations on Information Retrieval Performance. Proceedings of the 16th ACM-SIGIR Conf. on Research and Development in Information retrieval. p. 339-346. ACM Press, New York, NY, 1993.
Belkin et al. 95:
N.J.Belkin, P. Kantor, E.A. Fox and J.A. Shaw. Combining the Evidence of Multiple Query Representation for Information Retrieval. Information Processing and Management, 31:
431-448,1995.
Borlund and Ingwersen 97:
P. Borlund and P. Ingwersen. The development of a method for the evaluation of interac-tive information retrieval Systems. Journal ofDocumenfaf/on, 1997 (in press).
Brier96:
S. Brier. Cybersemiotics: a New Interdisciplinary Development Applied to the Problems of Knowledge Organization and Document Retrieval in Information Science. Journal of Documentation, 52(3): September, 1996 (in press).
Brookes 80:
B.C. Brookes. The Foundations of Information Science. Part 1: Philosophical Aspects.
Journal of Information Science, 2:125-133,1980.
Buckland 91:
M. Buckland. Information and Information Systems. Praeger, London, 1991.
Callan and Croft 93:
J.P. Callan and W.B. Croft. An Evaluation of Query Processing Strategies Using the TIP-STER Collection. In R. Korfhage, E.M. Rasmussen and P. Willett (eds). Proceedings of the Sixteenth Annual International Conference on Research and Development in Infor-mation Retrieval. p. 347-356. ACM, New York, 1993.
De Mey 80:
M. De Mey. The Relevance of the Cognitive Paradigm for Information Science. In. O. Har-bo, et. al. (eds.). Theory and Application of Information research. p. 48-61. Mansell, Lon-don, 1980.
Dervinand Nilan 86:
B. Dervin and M. Nilan. Information Needs and Uses. ARIST, (21):3-33,1986.
Ellis 89: D.
Ellis. A Behavioural Approach to Information Retrieval Systems Design. Journal of Docu-mentation, 45:171-212,1989.
Ellis 90: D.
Ellis. New Horizons in Information retrieval. Library Association, London, 1990.
Ellis 92:
D. Ellis. The physical and cognitive paradigms in information retrieval research. Journal of Documentation, 48(1): 45-64,1992.
Fox et al. 93:
E.A. Fox, D. Hix, L. Nowell, D. Bruneni, W. Wake, L. Heath and D. Rao. Users, User Inter-faces, and Objects: Envision, a Digital Library. Journal ofthe American Society for Infor-mation Science, 44: 480-491,1993.
Hancock-Beaulieu 92:
M. Hancock-Beaulieu. Query Expansion: Advances in Research in Online Catalogues.
Journal of Information Science, 18: 99-103,1992.
Hjorland and Albrechtsen 95:
B. Hjorland and H. Albrechtsen. Toward a New Horizon in Information Science: Domain Analysis. Journal of American Society for Information Science, 46(6,): 400-425, 1995.
Ingwersen 82:
P. Ingwersen. Search Procedures in the Library Analysed from the Cognitive Point of View. Journal of Documentation, 38:165-191,1982.
Ingwersen 92:
P. Ingwersen. Information Retrieval Interaction. Taylor Graham; London, 1992.
Ingwersen 93:
P. Ingwersen. The Cognitive Viewpoint in IR. Journal of Documentation, 49(1): 60-64, 1993.
Ingwersen 94:
P. Ingwersen. Information Scxience as a Cognitive Science. In H. Best, B. Endres-Nigge-meyer, M. herfurth and H.P. Ohly (eds.). Informations- und Wissenverarbeitung in den Sozialwissenschaften, p. 23-56. Westdeutscher Verlag, Opladen, Germany, 1994.
Ingwersen 96:
P. Ingwersen. Cognitive Perspectives of Information Retrieval Interaction: Elements of a Cognitive IR Theory. Journal of Documentation, 51 (1): 3-50,1996.
Ingwersen and Willett 95:
P. Ingwersen and P. Willett, P. An Introduction to Algorithmic and Cognitive Approaches for Information Retrieval. Libri, 45: 160-177,1995.
Katzer 82:
J. Katzer. A Study of the Overlap among Document Representations. Information Tech-nology: Research and Development, 2: 261 -274,1982.
Kuhlthau 93:
C. Kuhlthau. Seeking Meaning. Ablex Publ., New York, 1993.
Lancaster91:
F.W. Lancaster. Indexing in Theory and Practice. Library Association, London, 1991.
McCain 89:
K.W. McCain. Descriptor and Citation Retrieval in the Medicine Behavioural Sciences
Li-terature: Retrieval Overlaps and Novelty Distribution. Journal of American Society for In-formation Science, 40:110-114, 1989.
Pao 94:
M.L. Pao. Relervance Odds of retrieval Overlaps f rom Seven Search Fields. Information Processing & Management, 30(3): 305-314, 1994.
Peirce1891:
C.S. Peirce. The Architechtures of Theories. The Monist, 1 (2), January 1891.
Rasmussen et al.:
J. Rasmussen, A.M. Pejtersen and L.P. Goodstein. Cognitive Engineering: Concepts and Applications. Wiley, New York, 1992.
Salton 89:
G. Salton. Automatic Text Processing. The Transformation, Analysis and Retrieval of In-formation by Compufer.Addison-Wesley, Reading, Mass., 1989.
Salton et al. 94:
G. Salton, J. Allan and C. Buckley. Automatic Structuring and Retrieval of Large Text Files. Communications ofthe ACM, 37(2): 97-108, 1994.
Schamberetal. 90:
L.Schamber, M. Eisenberg and M. Nilan. A Re-examination of Relevance: Toward a Dy-namic, Situational Defintion. Information Processing & Managment, 26(6):755-776, 1990.
Shannon and Weaver 49:
C.E. Shannon and W. Weaver. The Mathematical Theory of Communication. University of Illinois Press, Urbana, III., 1949.
Su94:
L. Su. The Relevance of Recall and Precision in User Evaluation". Journal of the Ameri-can Society for Information Science, 45: 207-217,1994.
Tenopir89:
C. Tenopir. Issues in Online Database Searching. Libraries Unlimited, Englewood Cliffs, NJ, 1989.
Turtle and Croft 90:
H. Turtle and W.B. Croft. Inference Methods for Document Retrieval. In: J. Vidick (ed).
ACM-Sigir Conf. Proc. pp. 1-24. Bruxelles University, Bruxelles, 1990.
van Rijsbergen and Laimas 96:
C.J. van Rijsbergen and M. Laimas. An Information Calculus for Information Retrieval.
Journal of American Society for Information Science, AI'(5): 385-398,1996.
Wood-Harper et al. 85:
A.T. Wood-Harper, L. Antill, D.E.Avison. Information Systems Definition: The Multiview Approach. Blackwell Scientific Publ., Oxford, 1985.