Facet-based Exploratory Search in Topic Maps

26
Facet-based Exploratory Search in Topic Maps M. Ueberall, O. Drobnik Introduction Metrics Prototype Comparison Summary Facet-based Exploratory Search in Topic Maps M. Ueberall , O. Drobnik Telematics Group, Institute of Computer Science Goethe-University, Frankfurt/Main, Germany 2008 / 10 / 16

description

Ueberall, M., Drobnik, O.: Facet-based Exploratory Search in Topic Maps in L. Maicher, L.M. Garshol (Eds.): Proc. 4th International Conference on Topic Maps Research and Applications (TMRA 2008), Leipziger Beiträge zur Informatik: XII 49–62, ISBN 978-3-941152-05-2

Transcript of Facet-based Exploratory Search in Topic Maps

Page 1: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Facet-based Exploratory Searchin Topic Maps

M. Ueberall, O. Drobnik

Telematics Group, Institute of Computer ScienceGoethe-University, Frankfurt/Main, Germany

2008 / 10 / 16

Page 2: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Motivation and Objective

• The majority of Topic Maps based applications usesfaceted classification

• Still, generic exploratory search interfaces for TopicMaps which shield the user from representationaldetails have hardly been discussed

• In the following, we• extend an RDF based approach [Delbru et al. 2006]• discuss combinations with existing user interfaces

Page 3: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Exploratory Search

• Exploratory Search addresses information-seekingproblems where a user needs to find out somethingabout a domain but lacks specific (a-priori) knowledge

• In this situation, the user will usually• submit tentative queries• explore the retrieved information

in order to selectively seek and passively obtain cluesabout his next steps

Page 4: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Faceted Classification

• A Faceted Classification System enables• the assignment of multiple classifications (called facets)

to an object• the flexible ordering of these classifications in multiple

ways w/o following pre-determined, taxonomic order

• A facet is a metadata attribute which should represent asingle important characteristic of the classified objects

Page 5: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Facet-based Navigation asdecision tree traversal

Italy ...Country

... Lucca

Contains

Puccini, Giacomo Catalani, Alfredo

Birthplace of

Tosca ...Composed

Scenes de la

vie de Boheme ...Novel

La Boheme

(Leoncavallo)

La Boheme

(Puccini)

Source of

Giacomo, Guiseppe Illica, Luigi

Libretto by

Illica, Luigi ...Wrote libretto for

By iteratively choosing a facet (and associated restrictionvalues), the information space is traversed

Page 6: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Automated Facet Identification

• In the context of heterogenous, dynamically changingdatasets new information has to be (re-)classifiedon-the-fly

• In order to simplify the problem of missing/outdatedfacet-based classifications, a generic heuristic isneeded

Page 7: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Facet IdentificationEntity

http://company.com

http://product.com

Company

http://www.frankfurt.de

Person Consultant/Programmer

(Location) (Location)

Employment(Employs/Employed by)

Employer Employee

Person entity

(website)

(website) (job)

An entity is a subgraph of an information space, extractedby taking all adjacent vertices (objects) of a given vertex(subject)

Page 8: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Facet IdentificationView

http://company.com

http://product.com

Company

http://www.frankfurt.de

Person Consultant/Programmer

(Location) (Location)

Employment(Employs/Employed by)

Employer Employee

Company entityPerson entity

(website)

(website) (job)

A view is a set of entities of an information space

Page 9: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Facet IdentificationFacet

http://company.com

http://product.com

Company

http://www.frankfurt.de

Person Consultant/Programmer

(Location) (Location)

Employment

(Employs/Employed by)

Employer Employee

Company entityPerson entity

(website)

(website) (job)

A facet is a set of equally labeled edges in a view

Page 10: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Facet IdentificationRestriction Value

http://company.com

http://product.com

Company

http://www.frankfurt.de

Person Consultant/Programmer

(Location) (Location)

Employment(Employs/Employed by)

Employer Employee

Company entityPerson entity

(website)

(website) (job)

The set of objects connected to a facet represent therestriction values

Page 11: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Navigation MetricsBalance

Italy ...Country

... Lucca

Contains

Puccini, Giacomo Catalani, Alfredo

Birthplace of

Tosca ...Composed

Scenes de la

vie de Boheme ...Novel

La Boheme

(Leoncavallo)

La Boheme

(Puccini)

Source of

Giacomo, Guiseppe Illica, Luigi

Libretto by

Illica, Luigi ...Wrote libretto for

Idea:• the balance of a facet indicates its na-

vigation efficiency (cf. tree traversal)

Computation:• determine the (non-linear) normalised variance of the

number of subjects for each object

Page 12: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Navigation MetricsCardinality

Idea:• a suitable facet has a limited amount of restriction

values to choose from

Computation:• determine the number of different objects (restriction

values) for each facet• normalise result using a function based on the

gaussian density (cf. bell-shaped curve)

Page 13: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Navigation MetricsFrequency

Idea:• suitable facets occur frequently inside the collection• the more distinct concepts are covered, the more useful

the respective facet is in dividing the information space

Computation:• determine the number of subjects in the dataset for

which the facet has been defined• normalise result as a fraction of the total number of

subjects

Page 14: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Navigation MetricsExample (1)

http://company.com

http://product.com

Company

http://www.frankfurt.de

Person Consultant/Programmer

(Location) (Location)

Employment(Employs/Employed by)

Employer Employee

Company entityPerson entity

(website)

(website) (job)

facet balance(f) card(f) freq(f) scoreEmployment 1.0 0.72615 1.0 0.72615(website) 1.0 0.72615 0.5 0.36308(job) 1.0 0.0 0.5 0.0(location) 1.0 0.0 1.0 0.0

Page 15: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Navigation MetricsExample (2)

http://company.com

http://product.com

Company

http://www.frankfurt.de

Person Consultant/Programmer

(Location) (Location)

Employment(Employs/Employed by)

Employer Employee

(website)

(website) (job)

facet balance(f) card(f) freq(f) scoreEmployment 1.0 0.72615 0.33333 0.24205(website) 0.9 0.78270 0.5 0.35222(job) 1.0 0.72615 0.33333 0.24205(location) 0.9 0.78270 0.5 0.35222

Page 16: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Additional Facet Classes

• Basic requirements of a facet browser:• present the instances of all available types• the relations need to be made explicit, selectable

• Previous example focused on relations between topics:• association types• occurrence types

• Exploratory Search is also about filtering:• topic types• association roles types• scope (scoping topics)

Page 17: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Prototype

based on the /facet userinterface [Hildebrand et al.2006]

Objectives:

• support both TMAPI1 and TMAPI2 interfaces

• provide light-weight module for multiple query engines

Page 18: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Topic Map Exploration“Classic” generic views

Associations (18)

Born in

Lucca

Composed

La Bohème

Edgar

La fanciulla del West

Gianni Schicchi

Madama Butterfly

Manon Lescaut

La rondine

Suor Angelica

Il Tabarro

Tosca

Turandot

Le Villi

Died in

Brussels

Exponent of

verismo

Pupil of

Angeloni, Carlo

Bazzini, Antonio

Ponchielli, Amilcare

Internal Occurrences (4)

Bibliography

Budden, Julian: "Puccini: His Life and Works", Oxford University Press (Oxford, 2002)

Sadie, Stanley (ed): "Puccini and His Operas", Macmillan (London, 2000)

Date of birth

1858-12-22

Date of death

1924-11-29

External Occurrences (12)

Article

http://en.wikipedia.org/wiki/Giacomo_Puccini - Scope: Web; Wikipedia

http://localhost:8080/operamap/occurs/snl/puccini.htm - Scope: Local; Store Norske Leksikon

http://www.ontopia.net/topicmaps/examples/opera/occurs/snl/puccini.htm - Scope: Store Norske Leksikon; Web

Gallery

http://localhost:8080/operamap/occurs/puccini-gallery.htm - Scope: Local

Illustration

http://localhost:8080/operamap/occurs/composer/puccini.gif - Scope: Local

Sound clip

http://www.puccini.it/files/vocepucc.wav - Scope: Centro studi Giacomo Puccini; Italian; Web

Web page

http://localhost:8080/operamap/occurs/hnh-puccini.htm - Scope: Local; Naxos

http://www.naxos.com/composer/btm.asp?fullname=Puccini,%20Giacomo - Scope: Naxos; Web

http://www.operone.de/komponist/puccini.html - Scope: Web

http://www.r-ds.com/opera/pucciniana/gallery.htm - Scope: OperaResource; Web

Web site

http://www.landofpuccini.com

http://www.puccini.it - Scope: Centro studi Giacomo Puccini; Italian; Web(a) (b)

Page 19: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Topic Map ExplorationStylized, facet-based display

Puccini, Giacomo

La Boheme

Edgar

La fanciulla del West

Gianni Schicchi

Madame Butterfly

...

http://www.landofpuccini.com

http://www.puccini.it

http://en.wikipedia/. . .

http://localhost:8080/. . .

http://www.ontopia.net/. . .

http://localhost:8080/. . .

http://www.noxos.com/. . .

http://www.r-ds.com/. . .

http://www.operone.de/. . .Budden, Julian: “Puccini. . .

Sadie, Stanley (ed): “Puccini. . .

Composed by

(Composed)

Work Composer

Web site

Article

Web page

Bibliography

score=0.98012

score=0.72614

score=0.78270

score=0.83527

score=0.72614

Page 20: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Topic Map ExplorationSelection tree and resulting tolog query

$X

Giacosa, Guiseppe

...

... ... Italy

Libretto by

Audio recording

Composed by

Born in Located in

$A

$B $C

Query:

libretto-by($X : opera, giacosa : librettist),

audio-recording($X, $A),

composed-by( $X : work, $B : composer),born-in( $B : person, $C : place),located-in($C : containee, italy : container)?

A B C X

5 56301 2 Puccini, Giacomo Lucca Manon Lescaut

OPD-1265 Catalani, Alfredo Lucca Loreley

423 567-2 Puccini, Giacomo Lucca Madama Butterfly

7 47175 8 Puccini, Giacomo Lucca Tosca

[Omnigator] Query results http://localhost:8080/omnigator/plugins/tolog/query.jsp?codeexample=exRecordings3&query=libretto...

1 von 1 15.10.2008 14:28

Page 21: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Perspectives

• Navigation process for large information spaces can beimproved by reducing the initial number of differentfacets associated with a view

• Additional metrics: concept of semantic distancebetween vertices [Andres/Naito2008]

• Minimal sub-graph of a given list of interested topics asstarting point [Dichev/Dicheva/Fischer2007]

Page 22: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Summary and Outlook

• The presented exploratory search interface for topicmaps may enhance existing navigation aids

• It can be used to construct queries just by followinglinks between concepts of interest, regardless of theunderlying query engine/legend

• A stand-alone, TMAPI1/TMAPI2 based implementationwith both basic text-based and graphical interfaces willbe available shortly

• Currently, the forementioned functionality is beingintegrated in the user interface of an Eclipse basedprototype for software engineering support

Page 23: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Thank you!

e-mail to:[email protected]

Page 24: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Example scenario in LTM notation

[employer = "Employer"][employee = "Employee"][employment = "Employment"

= "Employs" / employer = "Employed by" / employee]employment([person = "Person"] : employee,

[company = "Company"] : employer)[website = "Website"][location = "Location"][job = "Job"]{company, website, "http://company.com/"} ∼company-website{company, website, "http://product.com/"} ∼product-website{company, location, "http://www.frankfurt.de"} ∼company-location{person, location, "http://www.frankfurt.de"} ∼person-location{person, job, [[consultant/programmer]]} ∼person-job

Page 25: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

Merging heterogenous topic mapsBibliography example

[kcc_title = "The Knowledge-Creating Company"][kcc_year = "1995"][nonaka = "Ikujiro Nonaka"][takeuchi = "Hirotaka Takeuchi"]reference( kcc_title :title, book :bibtype, kcc_year :year,

nonaka :author, takeuchi :author )

[kcc :book = "The Knowledge-Creating Company"]{ kcc, author, [[Ikujiro Nonaka]] }{ kcc, author, [[Hirotaka Takeuchi]] }{ kcc, year, [[1995]] }

Page 26: Facet-based Exploratory Search in Topic Maps

Facet-basedExploratorySearch in

Topic Maps

M. Ueberall,O. Drobnik

Introduction

Metrics

Prototype

Comparison

Summary

How to combine different scopes?

• Problem of lack of formal semantics [Garshol2008]

• A user must get an explanation/decide how scopingtopics are handled (and/or problem)

• Concepts from views of different scope can be handledin two ways:

• list only valid concepts using "merged" scope• individual concepts retain their original scope

• Either way, facets representing scoping topics cannoteasily be handled like other classes of facets (i.e., usercannot be shielded from representational details)