a context-aware model for the analysis of user interaction and qoe in mobile environments

Post on 24-Jun-2015

382 Views

Category:

Career

2 Downloads

Preview:

Click to see full reader

DESCRIPTION

by Pedro Mateo (pedromateo@um.es)

TRANSCRIPT

A Context-aware Model for the Analysis of UserInteraction and QoE in Mobile Environments

Pedro Mateopedromateo@um.es

CENTRIC - November 2012

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 1 / 16

Agenda

A Context-aware Model for the Analysis of User Interaction andQoE in Mobile Environments:

1 Research Questions

motivation, questions and proposed solutions

2 Research Work

starting point, context and rating data, QoE applications, user profiling

3 Conclusion

advantages, disadvantages and research directions

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 2 / 16

IResearch questions

Motivation

We analyzed user-system interactionelaborately...

...but avoiding interaction context andusers feedback

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 4 / 16

Motivation

However...interaction is mobile → different contexts

users’/experts’ impressions mostly ignored→ we can not determine quality

Moreover...lack of uniform approach to represent interaction and its context

different representations in which to base QoE inference

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 5 / 16

Research questions and proposed solutions

Question 1:

Incorporate rating and context data into interaction analysis?⇒ rating/context metrics within a user interaction meta-model

Question 2:

Compare QoE of different users/systems/contexts?⇒ base QoE inference on a common representation of interaction

Question 3:

Build profiles directly from successful-QoE experiences?⇒ use multi-dimensional interaction representations

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 6 / 16

IIResearch work

The starting point

meta-model to quantify user/systeminteraction → multimodal contexts

interaction described by turn→ parameters annotated “step by step”

relationship (data ⇔ time)

uniform/common representation

¬ context∨

¬ ratings ⇒ ¬ QoE

More info: Mateo & Hillmann, 2012, www.catedrasaes.org/wiki/MIM

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 8 / 16

Incorporating context and rating data

Context parameters added by turn:

dynamic description of context

context history

data automatically extracted with,e.g., Android HCI Extractor∗

User ratings as the used inquestionaries, e.g., AttrackDiff ∗∗

extracted automatically??

∗Android HCI Extractor: Mateo, 2011, code.google.com/p/android-hci-extractor∗∗AttrakDiff: Hassenzahl et al., 2003, www.attrakdiff.de

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 9 / 16

QoE applications

different interaction sources ⇒ same representation

unified criteria to base QoE inference

relationship (interaction ⇔ context ⇔ ratings ⇔ time)

run-time and offline applications

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 10 / 16

User profiling

multi-dimensional representation of interaction

user/application profiling based on successful QoE experiences

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 11 / 16

IIIConclusion

Good stuff...

common metrics + uniform representation⇒ common basis to determine/analyze QoE in different scenarios

strong relationship (interaction ⇔ context ⇔ ratings ⇔ time)⇒ dynamic analysis of QoE⇒ interaction history (prognosis)

multi-dimensional user profiling

QoE-aware decisions at run-time

reuse of metrics and tools

Project info: Mateo et al., www.catedrasaes.org/wiki/Caim

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 13 / 16

Not so good stuff...

diversity of interaction/context parameters & data abstraction level

cognitive data? extract subjective data automatically?

base the process on current device capabilities

early stage of research

these limitations and pitfalls are the challenges ofthis research!!

Project info: Mateo et al., www.catedrasaes.org/wiki/Caim

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 14 / 16

And now what?

Keep on research

select useful context and rating attributes

look for efficient and realistic data sources and extraction methods

Implement this solution

extend the base interaction meta-model

extend the Android HCI Extractor

Integrate this solution in real projects

use the meta-model for real analysis and QoE inference

Project info: Mateo et al., www.catedrasaes.org/wiki/Caim

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 15 / 16

Acknowledgements

www.catedrasaes.org

Supported by:

cátedrasaeslabs

Pedro Mateo pedromateo@um.es () Context-aware Interaction Model CENTRIC - November 2012 16 / 16

top related