Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
Transcript of Momm 2013
Charles Gouin-Vallerand & Alberto Montero
Actual recommender systems for smart phones do not really manage the context of the users :
A user is going into a cafe (e.g. Starbucks coffee) and he must :
find with Google play (or in its tons of download apps) the cafe's app;
or scan the QR code to download the app.
A user is going in the public transportation and :
wonder what other users in the train are using as apps;
find the usual apps (e.g. games, mails, news) he his using among several downloaded apps.
A context-aware recommender system would allows, among others, to :
Recommend apps related to specific locations and context;
Propose apps related to the categories of location or to similar context;
Recommend apps based on other users behavior to a specific location or context;
Help HCI by proposing dowloaded apps that are often used in a specific context/location.
There was a number of related works (Levandoski et al. 2012, Adomavicius et al. 2005, Yu et al. 2006) proposing location or context-based recommendation, but :
they were mostly working with 1 recommendation modality at a time;
they didn't asked to potential users want they want and how;
Moreover, with smart cities, how can we use all the contextual information to help users
Goal of this work
: apply our context-awareness model to smart urban environment scenarios and extended our work on service provision to the service recommendation :
With a survey :
Ask to potential users about which contextual information are the most important for a context-aware recommendation system in smart urban environments;
Use the potential users knowledge to measure which recommendation modalities are the most important and how a future recommendation system should behave;
Midterm goal : Develop a context-aware recommender system based on these measured modalities
We create a survey on the Surveymonkey website :
We used Facebook, Twitter and a french mailing list on HCI to recruit participants
The survey asked :
mobile devices utilization;
five scenarios of service recommendation: in a restaurant, in public transportation, while they are shopping, at their home and during a trip/vacation.
For each scenario, the participant had to describe which recommendation modalities are the most important;
questions about mobile apps versus different context settings.
311 participants filled the survey ... 55 participants completed it entirely
potential cause of the dropout : Survey was pretty long and complex (in the survey comments);
there was no compensations.
Participant profiles :
37% were male / 63% female
Average age : 33 years old with std = 10
57% from France, 33% from Canada
36% are living in 800k+ pop. cities (70% in 100k+)
73% are smart phone owners
13% are frequent mobile apps users (several times per hour)
Participants rated low the "other users' apps usage" modality:
they think that acceding to other users’ usage history is a privacy violation. Hypothesis corroborated by the users’ comments.
However, it's one the most used modalities in recommender systems like the in Amazon or Netflix websites (collaborative and content filtering)
Participants rated high their own apps usage for recommending apps.
Most important modalities are :
User's apps history
Current Context (e.g. location, time, kind of place)
Design and conception of a context-aware framework to recommend services in smart urban environment based on the study result
Will integrate potential users in the different conception steps (e.g. GUI evaluation) and evaluation on the field;
Integrate and manage the mood of the users in the recommendation modalities (Alberto Montero Ph.D. thesis)
Make a specific version of the framework to recommend assistance apps to seniors (project with Université du Québec at Chicoutimi and University of Sherbrooke, both in Canada)
Revue de la littérature
In the last years, we worked on :
Modeling context-awareness for smart urban environments based on a micro & macro context model (Gouin-Vallerand et al. (1) 2013 & 2012)
A context-aware service provision middleware for smart environments based on user interaction modalities (Gouin-Vallerand et al. (2) 2013)
Thank you for your attention!
Any inquries :
Website : http://vallerand.me
Twitter : @GouinVallerand
Special thanks to :
Quebec Research Funds - Nature & Technologies
Université du Québec
Anind K. Dey from Carnegie Mellon University (Post-doc advisor)