information retrieval in context of digital libraries - or dl in context of ir
DESCRIPTION
Information Retrieval in Context of Digital Libraries - or DL in Context of IR. Peter Ingwersen Royal School of LIS Denmark – [email protected] http://www.db.d/pi. Agenda. Information Retrieval In Context of Information Behavior Laboratory Model = Digital Library approach? - PowerPoint PPT PresentationTRANSCRIPT
Information Retrieval in Context of Digital Libraries
- or DL in Context of IR
Peter IngwersenRoyal School of LIS
Denmark – [email protected]://www.db.d/pi
LIDA 2009LIDA 2009 22IngwersenIngwersen
AgendaAgenda
Information RetrievalIn Context of Information BehaviorLaboratory Model = Digital Library
approach?Integrated Model – roles of context
The social perspective
Challenges in IR / DL according to model
Conclusions
LIDA 2009LIDA 2009 33IngwersenIngwersen
Information RetrievalInformation Retrieval
The processes involved in the representation, storage, searching, finding, filtering, presentation and use of information relevant to a requirement for information desired by a human user (The Turn, 2005)
Interaction – Time dimension
LIDA 2009LIDA 2009 44IngwersenIngwersen
Information behaviour and IRInformation behaviour and IRT. Wilson´s Onion Model, 1999 - extended:
Seeking
IRIR
Job-relatedWork TasksInterests
Non-job-relatedTasks and InterestsDaily-life behavior
Information behaviour
InteractiveIR
Behaviour
LIDA 2009LIDA 2009 55IngwersenIngwersen
Information behaviour … and other Information behaviour … and other central concepts in Information central concepts in Information
StudiesStudies Information behaviour:
to create information – e.g., on the Net - blogs; also human indexing, including social tagging;
to produce publications – e.g., as publisher
to communicate – face-to-face; chat; e-mail
to manage information sources – e.g. KM; selectivity
IB IB and other central and other central conceptsconcepts … …
Information seeking (behaviour) Information behaviour with interest for
Information Information need exist – even muddled or
exploratorySearching information sources – e.g.
colleagues Information Retrieval (I)IR
Searching information space via systems – Digital Library & Assets (interactive IR)
Retrieval models; relevance feedback & ranking; query modification; auto indexing and weighting;
LIDA 2009LIDA 2009 66IngwersenIngwersen
LIDA 2009LIDA 2009 77IngwersenIngwersen
The Laboratory Model of IRThe Laboratory Model of IR(in the Cranfield-TREC Laboratory Research (in the Cranfield-TREC Laboratory Research
Framework)Framework)
Could just as well be a model for Digital Library development
Docu-ments
Represen-tation
Database
Searchrequest
Query
Matching
Represen-tation
QueryResult
QueryResult
PseudoRelevanceFeedback
The Lab IR Cave, with a VisitorThe Turn – Ingwersen & Järvelin, 2005
Docu-ments
Represen-tation
Database
Searchrequest
Query
Matching
Represen-tation
QueryResult
QueryResult
Context
LIDA 2009LIDA 2009 99IngwersenIngwersen
Simplistic model of (I)IRSimplistic model of (I)IR – – short-term short-term interaction – in contextinteraction – in context
Informationobjects
IT: EnginesLogics
Algorithms
InterfaceInformation
Seeker(s)
Org.
Cultural
RQuery
R = Request / Relevance feedback
Short-term IS&R & social interactionCognitive transformations and influence over time
Modification
Social
Interaction
Social Tagging
Recommender techniques
SocialContext
LIDA 2009LIDA 2009 1010IngwersenIngwersen
Ingwersen
Central Components of Interactive Central Components of Interactive IR – the basic IR – the basic Integrated Integrated
FrameworkFramework
Informationobjects
IT: EnginesLogics
Algorithms
InterfaceCognitiveActor(s)
(team)
Org.
Cultural
SocialContext
Informationobjects
IT: EnginesLogics
Algorithms
InterfaceCognitiveActor(s)
(team)
Org.
Cultural
SocialContext
The Lab./DL Framework In situ
recommendation
In situ tagging
LIDA 2009LIDA 2009 1111IngwersenIngwersen
Ingwersen
Integrated Framework and Relevance Integrated Framework and Relevance CriteriaCriteria
Docs
Repr
DB
Request
Query
Match
Repr
Result
A: Recall, precision, efficiency
B: Usability, Graded rel., CumGain; Quality of information/process
C: Quality work process/result; Graded R.
Work TaskSeeking Task
SeekingProcess
WorkProcess
Task Result
Seeking Result
EvaluationCriteria:
Work task context
Seeking context
IR context
Socio-organizational& cultural context
D: Socio-cognitive relevance; quality of work task result
LIDA 2009LIDA 2009 1212IngwersenIngwersen
Moving into ContextMoving into ContextStrength:
Involvement of TASK (work/search) and …
Processes for fulfillment of task and …Task result / outcome
Seeking and retrieval tasks influenced by work tasks
Pointing to novel relevance measuresTask fulfillment measures; socio-
cognitive relevance; social utility (tagging, visits, downloads …)
LIDA 2009LIDA 2009 1313IngwersenIngwersen
Challenges to IR/DLChallenges to IR/DL
“[If] we consider that unlike art IR is not there for its own sake … then IR is far, far more than a branch of computer science”
And what information and relevance means to IR, Tefko Saracevic states (1997, p. 17) …
“[In] broadest sense: Information is … that involves not only messages (first sense) that are cognitively processed (second sense), but also a context – a situation, task, problem-at-hand, the social horizon, … intentions …”
LIDA 2009LIDA 2009 1414IngwersenIngwersen
Challenges to IR/DL – 2Challenges to IR/DL – 2
Understanding actors’ goals, tasks intentions – in diversity of contextsJob-related knowledge enquiries Daily-life information explorative behaviorsEntertainment - or simply ‘meaning making’
Inference of goals, tasks, intentions from implicit evidence from interaction behaviorImplicit relevance feedback study examples
LIDA 2009LIDA 2009 1515IngwersenIngwersen
Challenges to IR/DL – 3Challenges to IR/DL – 3
Leading to finding out the best algorithmic models and solutions – not in themselves – but given understanding of characteristics of searcher goals, …
A lot of searching is undirected, vague, random, exploratory, muddled … (Skov, 2009)
A lot of tagging (and folksonomies) is randomly done - but can be filtered
LIDA 2009LIDA 2009 1616IngwersenIngwersen
Challenges to IR/DL – 4Challenges to IR/DL – 4Belkin, Nick. Belkin, Nick. Sigir ForumSigir Forum, 42(1), 2008: 47-54, 42(1), 2008: 47-54
Recommender systems and personalization are relying on a narrow conception, applying vague correlations between a current searcher’s situation and previous Dwell time on page;Click-throughViewed, rated or saved objects by other
searchersSearch profiles’ contents
To tailor the rank of search results Or to find ‘things alike’ (probably better)
LIDA 2009LIDA 2009 1717IngwersenIngwersen
Challenges to IR/DL – 5Challenges to IR/DL – 5
Which of the (personal) contextual features do we need to involve – incl. the IT context?
How to present retrieved and filtered documents?Zooming in/out – integrated searching of media &
document types: presentation form and relevance/usability:
Are interface issues solved by Google snippets and Microsoft’s detail-whole format?
Alternative (elaborated) evaluation methods for interaction design (IR/DL) are required
LIDA 2009LIDA 2009 1818IngwersenIngwersen
The Circle of Systemic/Social Contexts in The Circle of Systemic/Social Contexts in interaction design: Digital Libraries & interaction design: Digital Libraries &
(I)IR – actor as centre(I)IR – actor as centre
Inter-face
CognitiveActor(s)
(team)
Org.
Cultural
SocialContext
Info.Objects
IT
Social Interactio
n
IR Interactio
n
LIDA 2009LIDA 2009 1919IngwersenIngwersen
ConclusionsConclusions
IR and DL (or Digital Assets including museums and cultural heritage) face same challenges of addressing the Interactive nature of the information
processContexts – and their limitsEvaluation & research approachesNeed for combined efforts of IT and
behavior
LIDA 2009LIDA 2009 2020IngwersenIngwersen
Thank You!