recomment: towards critiquing-based recommendation with speech interaction

34
S C I E N C E P A S S I O N T E C H N O L O G Y www.tugraz.at Critiquing-based Recommendation with Speech Interaction Peter Grasch ([email protected]) Alexander Felfernig ([email protected]) Florian Reinfrank ([email protected]), Institute for Software Technology October 15, 2013

Upload: peter-grasch

Post on 10-May-2015

86 views

Category:

Technology


2 download

TRANSCRIPT

Page 1: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

S C I E N C E P A S S I O N T E C H N O L O G Y

www.tugraz.at

Critiquing-basedRecommendationwith Speech InteractionPeter Grasch ([email protected])Alexander Felfernig ([email protected])Florian Reinfrank ([email protected]),Institute for Software Technology

October 15, 2013

Page 2: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Speech Interaction in Recommender Systems

(Written) natural language input has shown promisein (Shimazu 2001; arnestal, 2004)

Constraint satisfaction using spoken languagepresented by Thompson et al in 2004

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 20132

Page 3: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Critiquing-based Recommender Systems

Pioneering work as early as 1984: M. Williams’RABBIT [Williams, 1984]

Seminal work by Burke et al: FindMe[Burke et al., 1997]

Continued, active research, especially in the areas ofadvanced critiques[McCarthy et al., 2004, Zhang and Pu, 2006] anduser modeling[Reilly et al., 2005a, McCarthy et al., 2010]

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 20133

Page 4: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Traditional Critiquing-based Recommender

Figure : QwikShop [Reilly et al., 2005b]

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 20134

Page 5: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Concept

Preference Model

Recommendation

RecommendationStrategy

Preference Elicitation

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 20135

Page 6: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Concept

Preference Model

Recommendation

RecommendationStrategy

Preference Elicitation

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 20136

Page 7: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Rationale

A speech-based natural language interface can allowmore expressive feedback, thus reducing sessionlength.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 20137

Page 8: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Recommendation Strategy

Incremental unit critiquing-based system[Burke, 2000, Reilly et al., 2005a]

Prior probability based on sales rank

No initial search, relaxed similarity constraint

Custom utility function

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 20138

Page 9: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Recommendation Strategy

P ′ ← {p ∈ P|p satisfies last given critique};maxUtility ← −∞; bestOffer ← rold ;for p ∈ P ′ do

thisUtility ←∞ ;for c ∈ C do

thisUtility ← thisUtility + (1− c.ageMaxAge ) ∗ c.utility(p) ;

endif thisUtility > maxUtility then

maxUtility ← thisUtility ; bestOffer ← p ;end

endreturn bestOfferPeter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 20139

Page 10: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Utility Function

Control rate of change: Implicit goalsdistance = distance(a.value, p[a.id ].value) ∗ r .direction;perfectDist = metaModifier ∗ 0.5;if critiqueViolated then

return −abs(distance − perfectDist);else

if distance < perfectDist thenreturn

√distance

perfectDist ;

elsereturn max(perfectDist − distance + 1, 0.0001);

endend

Algorithm 1: Schematic utility calculation.Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201310

Page 11: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Utility Function

Control rate of change: Implicit goals

Figure : Utility function of the critique x > 50.Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201311

Page 12: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Utility Function

Control rate of change: Implicit goals

Figure : Utility function of subsequent critiques.Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201312

Page 13: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Utility Function

Control rate of change: Implicit goals

Figure : Utility function of subsequent critiques.Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201313

Page 14: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

ReComment: Speech Processing

Speech recognition solution based on CMU SPHINXand Simon [sph, 2013, sim, 2013]

Adapted to recommender situation

Keyword parser

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201314

Page 15: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Experiment

Comparison with traditional interface

80 participants

Measuring:

Interaction cyclesPerceived recommendation qualityUsability (adapted SUS)

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201315

Page 16: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Experiment: Traditional User Interface

Figure : ReComment: Mouse-based user interface.Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201316

Page 17: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Experiment: Speech-based User Interface

Figure : ReComment: Speech-based user interface.Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201317

Page 18: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Experiment: Speech-based User Interface

SentenceI am looking for a camera with 12 megapixel anda weight of around 200 gram.This camera with the same properties justsmaller.An even smaller camera.Optical zoom of 14 times would be better.More optical zoom.[...]

Table : Sample user interaction session.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201318

Page 19: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Results: Feedback Strategies

Category CountDiscarded 49 (12.8%)Unit critique 329 (85.7%)Compound critique (2 attributes) 3 (0.8%)Compound critique (3 attributes) 2 (0.5%)Compound critique (5 attributes) 1 (0.3%)

Table : Types of used commands.

74 sentences (20 %) referred to explicit values.12 sentences (3 %) used modifiers.Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201319

Page 20: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Results: Speech Processing

03

10

24

0

5

10

15

20

25

1 2 3 4

"Rec

omm

ent u

nder

stan

dsm

y vo

ice

inpu

t"

Figure : Participants’ perception of the speech-recognitionaccuracy ([1, 4], higher is better).

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201320

Page 21: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Results: Usability

Figure : Usability evaluation (adapted SUS scores).Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201321

Page 22: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Results: Recommendation Quality

0

5

10

15

20

25

Mouse−based interface Speech−based interface

Score

1

2

3

4

Figure : User score of last recommended item ([1, 4],higher is better).

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201322

Page 23: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Results: Recommender Efficiency

34.5

47.7

79.0250

10

20

30

40

50

Mouse interface(mean)

Mouse interface(median)

Speech interface(mean)

Speech interface(median)

Ses

sion

leng

th (

cycl

es)

Figure : Session length (lower is better).

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201323

Page 24: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Conclusion

Spoken language recommender systemsare worth exploring!

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201324

Page 25: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Future Work

Explore more natural user interfaces

Advanced sentiment analysis

Use of prosodic features, timing information, etc. toinfer certainty, frustration, etc.

Compare different recommender systems (e.g.,constraint based approaches)

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201325

Page 26: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Thank you for your attention.

Q & A

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201326

Page 27: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Sources I

[sim, 2013] (2013).

About Simon — Simon.

http://simon.kde.org.

[sph, 2013] (2013).

CMU Sphinx - Speech Recognition Toolkit.

http://cmusphinx.sf.net.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201327

Page 28: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Sources II

[Burke, 2000] Burke, R. (2000).

Knowledge-based recommender systems.

In Encyclopedia of Library and InformationSystems. Marcel Dekker.

[Burke et al., 1997] Burke, R. D., Hammond, K. J.,and Yound, B. (1997).

The findme approach to assisted browsing.

IEEE Expert, 12(4):32–40.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201328

Page 29: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Sources III

[McCarthy et al., 2004] McCarthy, K., Reilly, J.,McGinty, L., and Smyth, B. (2004).

On the dynamic generation of compound critiquesin conversational recommender systems.

In Adaptive Hypermedia and Adaptive Web-BasedSystems, pages 176–184. Springer.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201329

Page 30: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Sources IV

[McCarthy et al., 2010] McCarthy, K., Salem, Y., andSmyth, B. (2010).

Experience-based critiquing: reusing critiquingexperiences to improve conversationalrecommendation.

In Case-Based Reasoning. Research andDevelopment, pages 480–494. Springer.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201330

Page 31: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Sources V

[Reilly et al., 2005a] Reilly, J., McCarthy, K., McGinty,L., and Smyth, B. (2005a).

Incremental critiquing.

Knowledge-Based Systems, 18(4):143–151.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201331

Page 32: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Sources VI

[Reilly et al., 2005b] Reilly, J., Smyth, B., McGinty, L.,and McCarthy, K. (2005b).

Critiquing with confidence.

In Case-Based Reasoning Research andDevelopment, pages 436–450. Springer.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201332

Page 33: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Sources VII

[Williams, 1984] Williams, M. D. (1984).

What makes rabbit run?

International Journal of Man-Machine Studies,21(4):333–352.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201333

Page 34: ReComment: Towards Critiquing-based Recommendation with Speech Interaction

www.tugraz.at

Sources VIII

[Zhang and Pu, 2006] Zhang, J. and Pu, P. (2006).

A comparative study of compound critiquegeneration in conversational recommendersystems.

In Adaptive Hypermedia and Adaptive Web-BasedSystems, pages 234–243. Springer.

Peter Grasch ([email protected]), Institute for Software TechnologyOctober 15, 201334