Exploring Similarity: Improving Product Search with Parallel Coordinates

12
HCI 2014 Developer, designer and user characteristics Exploring Similarity Improving Product Search with Parallel Coordinates WED, 25 JUNE 2014 Mandy Keck, Martin Herrmann, Andreas Both, Dana Henkens, Rainer Groh Chair of Media Design - Technische Universität Dresden This work has been supported by the European Union and the Free State Saxony through the European Regional Development Fund

description

presentation @ HCI International 2014 – Developer, designer and user characteristics 25 JUNE 2014 – Crete, Greece

Transcript of Exploring Similarity: Improving Product Search with Parallel Coordinates

Page 1: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 Developer, designer and user characteristics

Exploring SimilarityImproving Product Search with Parallel CoordinatesWED, 25 JUNE 2014

Mandy Keck, Martin Herrmann,

Andreas Both, Dana Henkens, Rainer Groh

Chair of Media Design - Technische Universität Dresden This work has been supported by the European Union and the Free State Saxony through the European Regional Development Fund

Page 2: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 25 June 2014 M. Keck, M. Herrmann, A. Both, D. Henkens, R. Groh 2 | 12

Introduction

_ Research project »VISEA - Visual Search Interfaces«_ focus: visual interfaces for product search

PROBLEM:_ in the context of financial data, complex search masks often overstrain non-experts_ tables not suited for a quick comparison of multi- dimensional data

OUR CONCEPT:_ allows analysis of multi-dimensional data and comparison of various attributes at a glance_ enables the comparison and exploration of similar products

Page 3: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 25 June 2014 M. Keck, M. Herrmann, A. Both, D. Henkens, R. Groh 3 | 12

Search Paradigms

FACETED BROWSING:_ allows multiple access points for the search_ iterative refinement of the result set_ Faceted Classification describes items through a combination of facets_ each facet addresses a different property

SEARCH BY EXAMPLE:_ starts with an example_ can be generalized to find similar examples

Facet 1 Facet 2

facet value 1

facet value 2

facet value 3

facet value 4

facet value 1

facet value 2

facet value 3

Page 4: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 25 June 2014 M. Keck, M. Herrmann, A. Both, D. Henkens, R. Groh 4 | 12

Parallel Coordinates

Facet 1 Facet 2 Facet 3_ 2-dimensional representation of multi-dimensional data across a set of parallel axes_ each axis represents one attribute of the multi- dimensional data set_ a polyline represents a single data item_ well suited for continuous data attributes

Page 5: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 25 June 2014 M. Keck, M. Herrmann, A. Both, D. Henkens, R. Groh 5 | 12

Parallel Sets

Facet 1

Facet 2

_ basic layout is derived from parallel coordinates_ axes replaced by containers representing categories_ containers are scaled according to the frequency of the corresponding category_ containers are connected by streams representing the logical conjunction of the adjacent containers_ size of these streams give an impression of the frequency of items included in the conjunction_ mostly used for categorical data attributes

Page 6: Exploring Similarity: Improving Product Search with Parallel Coordinates

Interface Concept (BASIS LAYOUT)

Discount-Optionsschein DAX 20/12/2013 (SG31HF)

Bonus-Zerti�kat Munich Re 12/2014 (SG2103)

Outperformance-Zerti�kat Metro AG 19/06/2015 (SG3U5D)

Bonus-Zerti�kat BMW AG 12/2014 (SG2FPU)

Alpha-Zerti�kat S & P GSCU WTI ER (long) (SG12RH)

Discount-Optionsschein DAX 20/12/2013 (SG31HF)

Bonus-Zerti�kat Munich Re 12/2014 (SG2103)

Outperformance-Zerti�kat Metro AG 19/06/2015 (SG3U5D)

Bonus-Zerti�kat BMW AG 12/2014 (SG2FPU)

Alpha-Zerti�kat S & P GSCU WTI ER (long) (SG12RH)

Discount-Optionsschein DAX 20/12/2013 (SG31HF)

Bonus-Zerti�kat Munich Re 12/2014 (SG2103)

Outperformance-Zerti�kat Metro AG 19/06/2015 (SG3U5D)

Bonus-Zerti�kat BMW AG 12/2014 (SG2FPU)

Alpha-Zerti�kat S & P GSCU WTI ER (long) (SG12RH)

Discount-Optionsschein DAX 20/12/2013 (SG31HF)

Bonus-Zerti�kat Munich Re 12/2014 (SG2103)

Outperformance-Zerti�kat Metro AG 19/06/2015 (SG3U5D)

Bonus-Zerti�kat BMW AG 12/2014 (SG2FPU)

Alpha-Zerti�kat S & P GSCU WTI ER (long) (SG12RH)

Discount-Optionsschein DAX 20/12/2013 (SG31HF)

Bonus-Zerti�kat Munich Re 12/2014 (SG2103)

Outperformance-Zerti�kat Metro AG 19/06/2015 (SG3U5D)

RESULTS HISTORY COLLECTIONFACET 1 FACET 2 FACET 3

Produkttyp : Discount-Zerti�katBasiswerttyp : AktiePerformance : 0.28Anlagezeitraum : 1.4 JahreAusgabetag : 2012/8/8Bewertungstag : 2013/12/23Spread : 0.0006Link : ISIN: DE000SG3EHM7

comparebookmark

Discount-Operationsschein

Discount-Zerti�kat

Outperformance-Zerti�kat

Alpha-Zerti�kat

Bonus-Zerti�kat

Closed End Turbos

Open End Turbos

Zins

Rohsto�

Condor-Operationsschein

Währung

Aktie

Index

Volatilität

1– 6

0.5 – 1

0 – 0.5

-0.5 – 0

-1– -0.5

Result ViewParallel Axis View

Facet Valueheight shows the distribution of data items within the facets

Semantic Zoommouse scrolling zooms selected axisto explore the underlying hierarchy of a facet

Facet FilterSelecting a container activates a filter, leaving only items matching the facet value

Rearrangement of Axesallows a better comparison of two axes of interest

Result ListBrushing visually links a product on the left side to the associated list item

Historyitem can be added to the history,is highlighted in both views to distinguish them from unexplored or new products

Collectionproduct can be bookmarked and storedin a separate collection.

Detailed Product Information

Page 7: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 25 June 2014 M. Keck, M. Herrmann, A. Both, D. Henkens, R. Groh 7 | 12

Interface Concept (PARALLEL COORDINATES)

Faceted BrowsingMultiple filters can be activated on one or different axes to create complex filters

Search by examplecomparing feature can be used to identify products with similar properties,automatically centers all intersections

Fuzzy Filterscalable rectangular zone around the median line of the visualization,filter adjusts the opacity of the lines depending on a calculated relevance value

Page 8: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 25 June 2014 M. Keck, M. Herrmann, A. Both, D. Henkens, R. Groh 8 | 12

Interface Concepts (PARALLEL SETS)

Faceted BrowsingMultiple filters can be activated on one or different axes to create complex filters

Mouse-overhighlights all subsets, which contain the same items of the current subset

Page 9: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 25 June 2014 M. Keck, M. Herrmann, A. Both, D. Henkens, R. Groh 9 | 12

Demonstration

Page 10: Exploring Similarity: Improving Product Search with Parallel Coordinates

Evaluation (PARALLEL COORDINATES VS PARALLEL SETS)TIME IN SEC

ANALYSIS COMPARISON SEARCH

90

80

70

60

50

40

30

20

10

0

SOLUTION TIME FOR ALL TASKS

PARALLEL COORDINATES PARALLEL SETS

RESULTS OF THE QUESTIONNAIRE

4.5

4

3.5

3

2.5

2

1.5

1

0.5

0

LIKERT SCALE

LIKERT SCALE:5 - STRONGLY AGREE4 - AGREE3 - NEITHER AGREE OR DISAGREE 2 - DISAGREE1 - STRONGLY DISAGREE

A B C D E F

A - EFFECTIVENESS B - LEARNABILITYC - SATISFACTIOND - JOY OF USE E - EFFICIENCYF - RANGE OF FUNCTIONS

_ evaluation of both interface concepts for three different tasks analysis comparison search_ 13 users (7 females) in the age range of 23 – 60 years_ most of them were not familiar with PC/PS or financial products_ first part: participants had to solve 9 tasks per interface, divided into 3 task types with 3 tasks per type_ second part: participants had to complete a question- naire to evaluate each interface concept regarding effectiveness, learnability, satisfaction, joy of use, efficiency, and range of functions_ use of 5-point Likert scale.

Page 11: Exploring Similarity: Improving Product Search with Parallel Coordinates

Evaluation (FEATURES) THE FEATURE WAS USEFUL FOR SOLVING THE GIVEN TASKS

THE FEATURE WAS EASY TO USE

FEATURES

4.5

4

3.5

3

2.5

2

1.5

1

0.5

0

LIKERT SCALE

LIKERT SCALE:5 - STRONGLY AGREE4 - AGREE3 - NEITHER AGREE OR DISAGREE 2 - DISAGREE1 - STRONGLY DISAGREE

A B C D E F

A - FACET FILTERB - ZONE FILTERC - CENTER PRODUCT FOR COMPARING D - REARRANGEMENT OF AXESE - PANNING OF AXESF - ZOOMING OF AXESG - RESULT LIST

G

_ third part: users had to complete a questionnaire which evaluated individual features: facet filter zooming & panning of all axes rearrangement of the axes result list, history and collection comparing feature fuzzy filteroffered by PC & PS regarding usefulness and usability_ use of 5-point Likert scale

Page 12: Exploring Similarity: Improving Product Search with Parallel Coordinates

HCI 2014 25 June 2014 M. Keck, M. Herrmann, A. Both, D. Henkens, R. Groh 12 | 12

Conclusion & Future Work

_ combines faceted search and search by example paradigm with the visualization method of Parallel Coordinates / Parallel Sets_ distinctive strategies for analysis, comparing, and searching in large multi-dimensional datasets_ preliminary user study: all participants easily solved tasks from all of the three task types_ evaluation with different use cases (e.g. travel search) and user groups

Fig. 01 Travel Search with Parallel Set