1 / 26

THEORETICAL ISSUES IN CONTEXT MODELLING

THEORETICAL ISSUES IN CONTEXT MODELLING. Kalervo Järvelin (kalervo.jarvelin@uta.fi) University of Tampere Finland. Talk Outline. Lab IR - traditional study designs and contributions Contexts of IR Modeling context - issues in modeling A cognitive approach to modeling

chaela
Download Presentation

THEORETICAL ISSUES IN CONTEXT MODELLING

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. THEORETICAL ISSUES IN CONTEXT MODELLING Kalervo Järvelin (kalervo.jarvelin@uta.fi) University of Tampere Finland

  2. Talk Outline • Lab IR - traditional study designs and contributions • Contexts of IR • Modeling context - issues in modeling • A cognitive approach to modeling • Consequent research designs • Conclusion - does it matter?

  3. The Lab IR Framework Docu- ments Search request Relevance assessment Represen- tation Represen- tation Database Query Matching Query Result Evaluation Result Recall base Evaluation

  4. IR Research Goals The most popular focus - it is CS, after all, isn’t it?! • Traditional goals: • (a) theoretically (~formally) understanding IR, • (b) empirically testing IR methods, and • (c) designing better IR systems • This can be seen in typical IR study designs - by populating the framework with variables

  5. Study types: experimental methodological constructive theoretical Experimental designs – IR evaluation focus areas with variables designs operate with these fringe areas controlled (no analytical variables) dependent variables: ER many hidden variables Lab IR Study Types / Designs

  6. LabIR: Hypos, Laws, Theories • The dependent variables - typically recall and precision - using various metrics • The independent variables typically are the use or non-use of various techniques • The controlled variables - collection, requests • Thus Lab IR is about the explanation of the variation of recall and precision through IR techniques applied

  7. LabIR: Contributions • IR technology • Theoretical growth • theory expansion, e.g., due to enrichment of indexing theory • greater analytical power through formal model building • improved empirical support through testing over several test collections • new research questions – within the given set of variables • Theoretical growth has been limited by not bringing in radically new concepts or relations • Cannot possibly analyze real world applicability

  8. Lab IR Tests • IR research typically considers only retrieval tasks which most often are: • (a) purely topical • (b) content-only • (c) well-defined • (d) static, and • (e) exhaustive This is like saying that no matter what your situation is, your needs always are purely topical, content-only, well-defined, static, and … … True, isn’t it?

  9. Small World?

  10. Alternatives? Psssst ... ... IR in Context

  11. Work Task Situation Context Augmenting Task Performance Tools Education/Training Knowledge Methodology Acquire Knowledge Create Remember/Use Information Seeking Documents Other Sources Education Docs Available Find Docs Information Systems Colleagues ... Retrieve Docs ... Colleagues Other Collections DocDB Web FactDB One possible context:Task PerformanceAugmentation through IR

  12. Socio-organizational& culturalcontext Work task context Seeking context Seeking Task Work Task Docs Request Seeking Process Work Process Repr Repr DB Query Match Result Task Result Seeking Result IR context Nested Context Frameworks • How much should one cover? • How to model what is covered?

  13. What to cover? • Clearly, current IR test designs are not representative of real world IR • In principle, • whatever has important relationships with the application of IR, should be covered • however, what has, is not well known • answering this requires theoretical modeling, operationalization and empirical testing • so, what is context?

  14. Context? What?! • Dervin (1997): • there is no term that is more often used, less often defined, and when defined defined so variously as context - it has become almost a ritualistic invocation • for some, context has the potential of being virtually anything that is not defined as the focus • for others, it is inextricable surround that denies all generalizations • there are endless lists of contextual factors • This is why we hate it - it is foreign to CS!

  15. Some Contextual Themes • Reality is discontinuous, knowledge partial • Context is not useful as an independent entity • Context requires a focus on process and relationships between product and process • Need to focus on multiple interdependencies • Context is a necessary source of meaning (Dervin 1997)

  16. How to model? - IR Context Dimensions Each dimension containing multiple variables • Work task dimension • Search task dimension • Actor dimension • Perceived work task dimension • Perceived search task • Document dimension • Algorithmic search engine dimension • Algorithmic interface dimension • Access and interaction dimension Much more detail in the forthcoming book by Peter&Kal

  17. http://www.springeronline.com/

  18. A Cognitive Model of IS&R Information Objects 6 Organisat. 5 Social & Cognitive Actor(s) Interfaces 3 1 2 4 7 Information Systems Cultural Context 8 =Cognitive transformation and influence = Interactive communication of cognitive structures

  19. Perceived Information objects Org. / Cultural Situation i.e. perceived task – need sources Perceived Interface Perceived context Perceived IT: Engines Algorithms Social Cognitive World Real Social & Physical World Information objects Org. Information seeker Interface Social Context IT: Engines Algorithms Cultural Context around and within

  20. Augmented IR Research Goals • Theoretically understanding the application of IR in / to context; not just engines but also their use in varying situations • Empirically describing / testing the application of IR in / to various contexts • Supporting the design of information systems, services and information management • This does not downgrade IR into Social Science but rather means more / better engineering

  21. New Research Settings • Extended laboratory experiments • What kinds of work tasks are typical in Domain X? • What kinds of search tasks do these generate? • What kinds of documents tend to be relevant? • IIR experiments • Lab approach but including the seeking actor • Research on interfaces / actors in context • interaction of actors and information system interfaces in diverse socio-organizational contexts

  22. Lab IR settings focused on domain/task variables and dimensions of IT and information objects Information objects Social Actor(s) Interface Task Org. Context IT Cultural

  23. Lab IR settings focusing on actor variables and dimensions of IT and information objects Information objects Social Work task perception Interface Task Org. Context Actor types Search task perception IT Cultural

  24. Conclusion • Context is endless and pervasive • its management requires perspective, goals and interests: what do we want to achieve?? • we may see multiple, nested contexts • Thus, what is IR about? As technology ... • developing systems • based on a technological imperative? • based on people’s communication / access needs? • As a discipline ... [vs. a technology] • understanding ... perhaps ... but what? C-n-c-l?

  25. Does context really matter? • We do not know really (if, or how much) • Some certainly wish / believe it does not • I don’t believe them until shown that • no matter what one’s situation is, one’s information needs are of just one type • no matter of what type information needs are, they are always best served in one already known way

  26. Any comments or questions? Thank you!

More Related