In this paper, we study if reusing Google+ profiles can provide reliable recommendations on Twitter to resolve the cold start problem. Next, we investigate the impact of giving different weights for aggregating user profiles from two OSNs and present that giving a higher weight to the targeted OSN profile for aggregation allows the best performance in the context of a personalized link recommender system. Finally, we propose a user modeling strategy which combines entity-and category-based user profiles using with a discounting strategy. Results show that our proposed strategy improves the quality of user modeling significantly compared to the baseline method.
Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...
UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ and Twitter for Personalized Link Recommendations
1. Analyzing Aggregated Semantics-enabled
User Modeling on Google+ and Twitter for
Personalized Link Recommendations
Guangyuan Piao, John G. Breslin
Unit for Social
Semantics
24th Conference on User Modeling, Adaptation and Personalization
Halifax, Canada, 13-16, July, 2016
2. 2
Help me to tackle
the information
overload!
Recommend me
news articles that
now interest me!
Help me to find
interesting
(social) media!
Do not bother me with
advertisements that are
not interesting for me!
Give me personalized
support when I do my
online training!
Who is this? What are his
personal demands? How
can we make him happy?
Personalize my Social
Web experience!
The Social Web
3. Background – User Modeling
content enrichment
analysis &
user modeling
interest profile
?
personalized content
recommendations
(How) can we infer
user interest profiles
that support the
content recommender?
3*source: Analyzing user modeling on Twitter for personalized news recommendations, UMAP’11
4. Background – User Modeling
Representation of User Interest
Bag of
Words
Topic
Modeling
Bag of
Concepts
users' interests are
represented as a
set of words
topics are formed by groups
of co-occurring words and
each document is treated
as a mixture of topics
users' interests are
represented as a
set of concepts
• can exploit background
knowledge about concepts
for interest propagation
• focus on words
• assumption: a single doc contains rich information
• cannot provide semantic relationships among words
6. Aggregated User Interest Profiles
Interest Propagation
Related Work
flickr
Category:Indie_rock
The_Black_Keys
Eagles_of_Death_Metal genre
genre
7 times more interests using category-based profiles
might be helpful in the context of recommender systems
deliciou
s
stumble
upon
twitter
face
book
Abel et al. [UMUAI’13] Orlandi et al. [SEMANTiCS’12]
same weight for each Online Social Network(OSN) profile
7. Aggregated User Interest Profiles
• to investigate if giving a higher weight to the targeted OSN for
aggregation improves profiles without aggregation more
significantly
Interest Propagation
• to study category-based user profiles in the context of
recommender systems on Twitter compared to entity-based ones
• to propose and evaluate a mixed approach using entity- and
category-based user interest profiles
7
Aim of Work
8. 8
User Modeling Framework
propagation strategy
using DBpedia
1. T(Cat)
2. T(CatDiscount)
3. Tonly+T(x)
User Profiles
Category:
Smartphones
… iPhone
0.12 … 0.08Concept
Frequency
entity-based
user profiles
normalization
9. Twitter & Google+ Dataset from about.me
• 429 active users using Google+ and Twitter
Experiment
• task: recommending 10 links (URLs)
• recommendation algorithm: cosine similarity
• ground truth links: 10 links shared via tweets
• candidate links: 5,165 links
9
Experiment Setup
used for user modeling ground truth
recommendation time
links (URLs)
10. Results - MRR
10
Aggregated User Interest Profiles
• GmTn : m & n denote the weights for Google+ & Twitter
• Tonly : entity-based Twitter profile without aggregation
a higher weight for the
targeted OSN profile provides
the best performance
11. Results - recall
11
Aggregated User Interest Profiles
• GmTn : m & n denote the weights for Google+ & Twitter
• Tonly : entity-based Twitter profile without aggregation
a higher weight for the
targeted OSN profile provides
the best performance
12. Category-based User Profiles
• T(Cat): replacing entities with the categories from DBpedia
applying the same weights
• T(CatDiscount): applies a discounting strategy for the extended
categories
Entity- and Category-based User Profiles
• Tonly+T(x): combines the entity- and category-based profiles
Interest Propagation
SP: sub-pages
SC: sub-categories
15. Conclusions & Future Work
Conclusions
• a higher weight for the targeted OSN for aggregation
• category-based user profiles does not outperform entity-based
user profiles
• mixed approach outperforms entity- or category-based user
profiles
Future Work
• in the near future, we plan to investigate different aspects of
DBpedia for interest propagation
15
16. 16
Thank you for your attention!
Guangyuan Piao
homepage: http://parklize.github.io
e-mail: guangyuan.piao@insight-centre.org
twitter: https://twitter.com/parklize
slideshare: http://www.slideshare.net/parklize
Editor's Notes
- focus on words, which cannot provide semantic information and relationships among them
- focus on words, which cannot provide semantic information and relationships among them
- focus on words, which cannot provide semantic information and relationships among them