SlideShare a Scribd company logo
Analyzing User Modeling on Twitter for
Personalized News Recommendations
Best Paper of Conference on User Modeling, Adaption and Personalization (UMAP’11)
Authors: Fabian Abel, Qi Gao, Geert-Jan Houben and Ke Tao
Unit for Social Software
Presenter: Guangyuan Piao
Reading Group, 29/09/2015
Contents
• Background & Related Work
• Design Options for User Modeling in
Online Social Networks for Recommendations
• Research Questions
• Dataset for the study
• Study for Research Questions 1,2
• Experiment for Research Question 3
2/21
3/21
User Modeling in Online Social Networks
for Recommendations
4/21
Related Work
Representation of a User Model
• bag-of-words, Chen et al. for recommending and ranking URLs
posted in Twitter messages
Study of hashtags in Twitter
• investigate the specificity, stability over time, Laniado & Mika
• temporal dynamics of hashtags, Laniado & Mika / Huang et al.
Enrichment of tweets
• Exploit metadata of research papers to enrich the semantics of
tweets for mapping tweets to conference talks, Rowe et al.
5/21
Design Options for User Modeling
Semantic
Enrichment
Profile Type
Time Constraint
1. tweet-based
2. further enrichment
1. hashtag-based
2. entity-based
3. topic-based
1. time period
2. temporal patterns
User Profile
Google … iPhone
0.09 … 0.08
P(u) =
link news
monitored
news pool
an entity-based profile
6/21
Research Questions
1. how does the semantic enrichment impact the characteristics
and quality of Twitter-based profiles?
2. how do (different types of) profiles evolve over time? Are there
any characteristic patterns?
3. how do the different user modeling strategies impact
personalization (personalized news recommendation) and
does the consideration of temporal patterns improve the
accuracy of the recommendations?
7/21
Dataset for the study
Collected from Twitter with more than
• 20,000 Twitter users
• 2 months
• 10,000,000 tweets
• 75,000 news articles
Sample dataset for study
• 1,619 users with at least 20 tweets & 1 tweet/month (2,316,204 tweets)
8/21
Research Question 1
how does the semantic enrichment impact
the characteristics and quality of Twitter-based profiles?
9/21
• tweet-only-based user modeling, fails to create profiles for 100 users
• by enrichment with entities and topics obtained from linked news
articles ➡️ a higher # of distinct concepts and variety for per profile
10/21
The impact of news-based enrichment
• entity-based and topic-based strategies have higher coverage
Comparison of different types of profiles
11/21
Research Question 2
how do (different types of) profiles evolve over time?
are there any characteristic patterns?
12/21
• profile difference is measured by d1-distance
• user profiles change over time: the older a profile the more it differs
from the current profile of the user
Profile difference over time
13/21
d1 = 2 iPad iPhone
Px(u) 1 0
Py(u) 0 1
• the difference of weekday and weekend profiles is higher than that of
other temporal patterns
Profile difference of weekday and weekend
14/21
• hashtag-based & entity-based profiles change most while the types of
entities people refer to (person, product etc.) do not differ strongly
Profile difference of weekday and weekend
15/21
how do the different user modeling strategies impact
personalization (personalized news recommendation)?
and does the consideration of temporal patterns
improve the accuracy of the recommendations?
Research Question 3
16/21
Evaluation Setup for News Recommendations
Main goal: analyze and compare the applicability of the different user
modeling strategies in the context of news recommendations.
Recommendation algorithm: cosine similarity between user profile
and tweets.
Evaluation Metrics:
• MRR (Mean Reciprocal Rank): at which rank the first item relevant to
the user occurs on average.
• S@k (Success at rank k): the mean probability that a relevant item
occurs within the top k of the ranking, k=10.
observation time period for Twitter
last week
used for user modeling
get candidate
news items
17/21
Results for News Recommendations – impact of
news-based enrichment
• entity-based strategy is better than others
• exploiting both tweets and linked news articles for creating user
profiles improves the performance significantly (p < 0.05)
18/21
Results for News Recommendations – impact of
considering temporal patterns
• fresh user profiles for topic-based user modeling are more applicable
for recommending news articles, while complete user profiles for
entity-based user modeling yields better recommendations.
• similar patterns can be found for weekend news recommendations.
(fresh profile: two weeks before recommendation time)
19/21
Conclusions
Further enrichment with semantics extracted from news articles
• enhanced the variety of the constructed profiles
• improved the accuracy of news article recommendations
Temporal dynamics of user profiles
• user profiles change over time
• user vary from weekdays and weekends
• consideration of temporal dynamics is beneficial to news
recommendations for topic-based user modeling strategy
20/21

More Related Content

Similar to Analyzing User Modeling on Twitter for Personalized News Recommendations

SEMANTiCS2016 - Exploring Dynamics and Semantics of User Interests for User ...
SEMANTiCS2016 - Exploring Dynamics and Semantics of User Interests for User ...SEMANTiCS2016 - Exploring Dynamics and Semantics of User Interests for User ...
SEMANTiCS2016 - Exploring Dynamics and Semantics of User Interests for User ...
GUANGYUAN PIAO
 
Twitter in Academic Conferences
Twitter in Academic ConferencesTwitter in Academic Conferences
Twitter in Academic Conferences
Denis Parra Santander
 
From Tweetations to Citations: Social Media and the Researcher
From Tweetations to Citations: Social Media and the ResearcherFrom Tweetations to Citations: Social Media and the Researcher
From Tweetations to Citations: Social Media and the Researcher
Sharon Karasmanis
 
Enhancing Information Retrieval by Personalization Techniques
Enhancing Information Retrieval by Personalization TechniquesEnhancing Information Retrieval by Personalization Techniques
Enhancing Information Retrieval by Personalization Techniques
veningstonk
 
Lecture 5: How to make the Social Web Personalized? (VU Amsterdam Social Web ...
Lecture 5: How to make the Social Web Personalized? (VU Amsterdam Social Web ...Lecture 5: How to make the Social Web Personalized? (VU Amsterdam Social Web ...
Lecture 5: How to make the Social Web Personalized? (VU Amsterdam Social Web ...
Lora Aroyo
 
New Methodologies for Capturing and Working with Publicly Available Twitter Data
New Methodologies for Capturing and Working with Publicly Available Twitter DataNew Methodologies for Capturing and Working with Publicly Available Twitter Data
New Methodologies for Capturing and Working with Publicly Available Twitter Data
Axel Bruns
 
Twitter as a personalizable information service ii
Twitter as a personalizable information service iiTwitter as a personalizable information service ii
Twitter as a personalizable information service ii
Kan-Han (John) Lu
 
Slides ecir2016
Slides ecir2016Slides ecir2016
Slides ecir2016
Fattane Zarrinkalam
 
EKAW2016 - Interest Representation, Enrichment, Dynamics, and Propagation: A ...
EKAW2016 - Interest Representation, Enrichment, Dynamics, and Propagation: A ...EKAW2016 - Interest Representation, Enrichment, Dynamics, and Propagation: A ...
EKAW2016 - Interest Representation, Enrichment, Dynamics, and Propagation: A ...
GUANGYUAN PIAO
 
MyPlan - similarity metrics for matching lifelong learner timelines
MyPlan - similarity metrics for matching lifelong learner timelinesMyPlan - similarity metrics for matching lifelong learner timelines
MyPlan - similarity metrics for matching lifelong learner timelines
Nicolas Van Labeke
 
Research projects devyani jain
Research projects devyani jainResearch projects devyani jain
Research projects devyani jain
Devyani Jain
 
Use of social media by scholarly authors
Use of social media by scholarly authorsUse of social media by scholarly authors
Use of social media by scholarly authors
Kudos
 
Using Twitter as a data source: An overview of ethical challenges
Using Twitter as a data source: An overview of ethical challengesUsing Twitter as a data source: An overview of ethical challenges
Using Twitter as a data source: An overview of ethical challenges
Dr Wasim Ahmed
 
Communities of attention' around journal papers: Who is tweeting about scient...
Communities of attention' around journal papers: Who is tweeting about scient...Communities of attention' around journal papers: Who is tweeting about scient...
Communities of attention' around journal papers: Who is tweeting about scient...
Stefanie Haustein
 
Exploring Generative Models of Tripartite Graphs for Recommendation in Social...
Exploring Generative Models of Tripartite Graphs for Recommendation in Social...Exploring Generative Models of Tripartite Graphs for Recommendation in Social...
Exploring Generative Models of Tripartite Graphs for Recommendation in Social...
Charalampos Chelmis
 
Practical Uses of Altmetrics - Hear librarians talk about how they use new me...
Practical Uses of Altmetrics - Hear librarians talk about how they use new me...Practical Uses of Altmetrics - Hear librarians talk about how they use new me...
Practical Uses of Altmetrics - Hear librarians talk about how they use new me...
plumanalytics
 
UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ an...
UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ an...UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ an...
UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ an...
GUANGYUAN PIAO
 
E017433538
E017433538E017433538
E017433538
IOSR Journals
 
The role of new information and communication technologies in information and...
The role of new information and communication technologies in information and...The role of new information and communication technologies in information and...
The role of new information and communication technologies in information and...
Christina Pikas
 
Online Forums vs. Social Networks: Two Case Studies to support eGovernment wi...
Online Forums vs. Social Networks: Two Case Studies to support eGovernment wi...Online Forums vs. Social Networks: Two Case Studies to support eGovernment wi...
Online Forums vs. Social Networks: Two Case Studies to support eGovernment wi...
Timo Wandhoefer
 

Similar to Analyzing User Modeling on Twitter for Personalized News Recommendations (20)

SEMANTiCS2016 - Exploring Dynamics and Semantics of User Interests for User ...
SEMANTiCS2016 - Exploring Dynamics and Semantics of User Interests for User ...SEMANTiCS2016 - Exploring Dynamics and Semantics of User Interests for User ...
SEMANTiCS2016 - Exploring Dynamics and Semantics of User Interests for User ...
 
Twitter in Academic Conferences
Twitter in Academic ConferencesTwitter in Academic Conferences
Twitter in Academic Conferences
 
From Tweetations to Citations: Social Media and the Researcher
From Tweetations to Citations: Social Media and the ResearcherFrom Tweetations to Citations: Social Media and the Researcher
From Tweetations to Citations: Social Media and the Researcher
 
Enhancing Information Retrieval by Personalization Techniques
Enhancing Information Retrieval by Personalization TechniquesEnhancing Information Retrieval by Personalization Techniques
Enhancing Information Retrieval by Personalization Techniques
 
Lecture 5: How to make the Social Web Personalized? (VU Amsterdam Social Web ...
Lecture 5: How to make the Social Web Personalized? (VU Amsterdam Social Web ...Lecture 5: How to make the Social Web Personalized? (VU Amsterdam Social Web ...
Lecture 5: How to make the Social Web Personalized? (VU Amsterdam Social Web ...
 
New Methodologies for Capturing and Working with Publicly Available Twitter Data
New Methodologies for Capturing and Working with Publicly Available Twitter DataNew Methodologies for Capturing and Working with Publicly Available Twitter Data
New Methodologies for Capturing and Working with Publicly Available Twitter Data
 
Twitter as a personalizable information service ii
Twitter as a personalizable information service iiTwitter as a personalizable information service ii
Twitter as a personalizable information service ii
 
Slides ecir2016
Slides ecir2016Slides ecir2016
Slides ecir2016
 
EKAW2016 - Interest Representation, Enrichment, Dynamics, and Propagation: A ...
EKAW2016 - Interest Representation, Enrichment, Dynamics, and Propagation: A ...EKAW2016 - Interest Representation, Enrichment, Dynamics, and Propagation: A ...
EKAW2016 - Interest Representation, Enrichment, Dynamics, and Propagation: A ...
 
MyPlan - similarity metrics for matching lifelong learner timelines
MyPlan - similarity metrics for matching lifelong learner timelinesMyPlan - similarity metrics for matching lifelong learner timelines
MyPlan - similarity metrics for matching lifelong learner timelines
 
Research projects devyani jain
Research projects devyani jainResearch projects devyani jain
Research projects devyani jain
 
Use of social media by scholarly authors
Use of social media by scholarly authorsUse of social media by scholarly authors
Use of social media by scholarly authors
 
Using Twitter as a data source: An overview of ethical challenges
Using Twitter as a data source: An overview of ethical challengesUsing Twitter as a data source: An overview of ethical challenges
Using Twitter as a data source: An overview of ethical challenges
 
Communities of attention' around journal papers: Who is tweeting about scient...
Communities of attention' around journal papers: Who is tweeting about scient...Communities of attention' around journal papers: Who is tweeting about scient...
Communities of attention' around journal papers: Who is tweeting about scient...
 
Exploring Generative Models of Tripartite Graphs for Recommendation in Social...
Exploring Generative Models of Tripartite Graphs for Recommendation in Social...Exploring Generative Models of Tripartite Graphs for Recommendation in Social...
Exploring Generative Models of Tripartite Graphs for Recommendation in Social...
 
Practical Uses of Altmetrics - Hear librarians talk about how they use new me...
Practical Uses of Altmetrics - Hear librarians talk about how they use new me...Practical Uses of Altmetrics - Hear librarians talk about how they use new me...
Practical Uses of Altmetrics - Hear librarians talk about how they use new me...
 
UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ an...
UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ an...UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ an...
UMAP2016 - Analyzing Aggregated Semantics-enabled User Modeling on Google+ an...
 
E017433538
E017433538E017433538
E017433538
 
The role of new information and communication technologies in information and...
The role of new information and communication technologies in information and...The role of new information and communication technologies in information and...
The role of new information and communication technologies in information and...
 
Online Forums vs. Social Networks: Two Case Studies to support eGovernment wi...
Online Forums vs. Social Networks: Two Case Studies to support eGovernment wi...Online Forums vs. Social Networks: Two Case Studies to support eGovernment wi...
Online Forums vs. Social Networks: Two Case Studies to support eGovernment wi...
 

More from GUANGYUAN PIAO

Env2Vec: Accelerating VNF Testing with Deep Learning
Env2Vec: Accelerating VNF Testing with Deep LearningEnv2Vec: Accelerating VNF Testing with Deep Learning
Env2Vec: Accelerating VNF Testing with Deep Learning
GUANGYUAN PIAO
 
Domain-Aware Sentiment Classification with GRUs and CNNs
Domain-Aware Sentiment Classification with GRUs and CNNsDomain-Aware Sentiment Classification with GRUs and CNNs
Domain-Aware Sentiment Classification with GRUs and CNNs
GUANGYUAN PIAO
 
A Study of the Similarities of Entity Embeddings Learned from Different Aspec...
A Study of the Similarities of Entity Embeddings Learned from Different Aspec...A Study of the Similarities of Entity Embeddings Learned from Different Aspec...
A Study of the Similarities of Entity Embeddings Learned from Different Aspec...
GUANGYUAN PIAO
 
Retweet Prediction with Attention-based Deep Neural Network
Retweet Prediction with Attention-based Deep Neural NetworkRetweet Prediction with Attention-based Deep Neural Network
Retweet Prediction with Attention-based Deep Neural Network
GUANGYUAN PIAO
 
WISE2017 - Factorization Machines Leveraging Lightweight Linked Open Data-ena...
WISE2017 - Factorization Machines Leveraging Lightweight Linked Open Data-ena...WISE2017 - Factorization Machines Leveraging Lightweight Linked Open Data-ena...
WISE2017 - Factorization Machines Leveraging Lightweight Linked Open Data-ena...
GUANGYUAN PIAO
 
Hypertext2017-Leveraging Followee List Memberships for Inferring User Interes...
Hypertext2017-Leveraging Followee List Memberships for Inferring User Interes...Hypertext2017-Leveraging Followee List Memberships for Inferring User Interes...
Hypertext2017-Leveraging Followee List Memberships for Inferring User Interes...
GUANGYUAN PIAO
 
ECIR2017-Inferring User Interests for Passive Users on Twitter by Leveraging ...
ECIR2017-Inferring User Interests for Passive Users on Twitter by Leveraging ...ECIR2017-Inferring User Interests for Passive Users on Twitter by Leveraging ...
ECIR2017-Inferring User Interests for Passive Users on Twitter by Leveraging ...
GUANGYUAN PIAO
 
UMAP2016EA - Analyzing MOOC Entries of Professionals on LinkedIn for User Mod...
UMAP2016EA - Analyzing MOOC Entries of Professionals on LinkedIn for User Mod...UMAP2016EA - Analyzing MOOC Entries of Professionals on LinkedIn for User Mod...
UMAP2016EA - Analyzing MOOC Entries of Professionals on LinkedIn for User Mod...
GUANGYUAN PIAO
 
RDFa Basics
RDFa BasicsRDFa Basics
RDFa Basics
GUANGYUAN PIAO
 
Owl 2.0 Overview
Owl 2.0 OverviewOwl 2.0 Overview
Owl 2.0 Overview
GUANGYUAN PIAO
 
OWL 2.0 Primer Part01
OWL 2.0 Primer Part01OWL 2.0 Primer Part01
OWL 2.0 Primer Part01
GUANGYUAN PIAO
 
OWL2.0 Primer Part02
OWL2.0 Primer Part02OWL2.0 Primer Part02
OWL2.0 Primer Part02
GUANGYUAN PIAO
 
Hdd industry
Hdd industryHdd industry
Hdd industry
GUANGYUAN PIAO
 

More from GUANGYUAN PIAO (13)

Env2Vec: Accelerating VNF Testing with Deep Learning
Env2Vec: Accelerating VNF Testing with Deep LearningEnv2Vec: Accelerating VNF Testing with Deep Learning
Env2Vec: Accelerating VNF Testing with Deep Learning
 
Domain-Aware Sentiment Classification with GRUs and CNNs
Domain-Aware Sentiment Classification with GRUs and CNNsDomain-Aware Sentiment Classification with GRUs and CNNs
Domain-Aware Sentiment Classification with GRUs and CNNs
 
A Study of the Similarities of Entity Embeddings Learned from Different Aspec...
A Study of the Similarities of Entity Embeddings Learned from Different Aspec...A Study of the Similarities of Entity Embeddings Learned from Different Aspec...
A Study of the Similarities of Entity Embeddings Learned from Different Aspec...
 
Retweet Prediction with Attention-based Deep Neural Network
Retweet Prediction with Attention-based Deep Neural NetworkRetweet Prediction with Attention-based Deep Neural Network
Retweet Prediction with Attention-based Deep Neural Network
 
WISE2017 - Factorization Machines Leveraging Lightweight Linked Open Data-ena...
WISE2017 - Factorization Machines Leveraging Lightweight Linked Open Data-ena...WISE2017 - Factorization Machines Leveraging Lightweight Linked Open Data-ena...
WISE2017 - Factorization Machines Leveraging Lightweight Linked Open Data-ena...
 
Hypertext2017-Leveraging Followee List Memberships for Inferring User Interes...
Hypertext2017-Leveraging Followee List Memberships for Inferring User Interes...Hypertext2017-Leveraging Followee List Memberships for Inferring User Interes...
Hypertext2017-Leveraging Followee List Memberships for Inferring User Interes...
 
ECIR2017-Inferring User Interests for Passive Users on Twitter by Leveraging ...
ECIR2017-Inferring User Interests for Passive Users on Twitter by Leveraging ...ECIR2017-Inferring User Interests for Passive Users on Twitter by Leveraging ...
ECIR2017-Inferring User Interests for Passive Users on Twitter by Leveraging ...
 
UMAP2016EA - Analyzing MOOC Entries of Professionals on LinkedIn for User Mod...
UMAP2016EA - Analyzing MOOC Entries of Professionals on LinkedIn for User Mod...UMAP2016EA - Analyzing MOOC Entries of Professionals on LinkedIn for User Mod...
UMAP2016EA - Analyzing MOOC Entries of Professionals on LinkedIn for User Mod...
 
RDFa Basics
RDFa BasicsRDFa Basics
RDFa Basics
 
Owl 2.0 Overview
Owl 2.0 OverviewOwl 2.0 Overview
Owl 2.0 Overview
 
OWL 2.0 Primer Part01
OWL 2.0 Primer Part01OWL 2.0 Primer Part01
OWL 2.0 Primer Part01
 
OWL2.0 Primer Part02
OWL2.0 Primer Part02OWL2.0 Primer Part02
OWL2.0 Primer Part02
 
Hdd industry
Hdd industryHdd industry
Hdd industry
 

Recently uploaded

2 Post harvest Physiology of Horticulture produce.pptx
2 Post harvest Physiology of Horticulture  produce.pptx2 Post harvest Physiology of Horticulture  produce.pptx
2 Post harvest Physiology of Horticulture produce.pptx
UmeshTimilsina1
 
Parkinson Disease & Anti-Parkinsonian Drugs.pptx
Parkinson Disease & Anti-Parkinsonian Drugs.pptxParkinson Disease & Anti-Parkinsonian Drugs.pptx
Parkinson Disease & Anti-Parkinsonian Drugs.pptx
AnujVishwakarma34
 
MATATAG CURRICULUM sample lesson exemplar.docx
MATATAG CURRICULUM sample lesson exemplar.docxMATATAG CURRICULUM sample lesson exemplar.docx
MATATAG CURRICULUM sample lesson exemplar.docx
yardenmendoza
 
View Inheritance in Odoo 17 - Odoo 17 Slides
View Inheritance in Odoo 17 - Odoo 17  SlidesView Inheritance in Odoo 17 - Odoo 17  Slides
View Inheritance in Odoo 17 - Odoo 17 Slides
Celine George
 
Benchmarking Sustainability: Neurosciences and AI Tech Research in Macau - Ke...
Benchmarking Sustainability: Neurosciences and AI Tech Research in Macau - Ke...Benchmarking Sustainability: Neurosciences and AI Tech Research in Macau - Ke...
Benchmarking Sustainability: Neurosciences and AI Tech Research in Macau - Ke...
Alvaro Barbosa
 
Odoo 17 Events - Attendees List Scanning
Odoo 17 Events - Attendees List ScanningOdoo 17 Events - Attendees List Scanning
Odoo 17 Events - Attendees List Scanning
Celine George
 
MVC Interview Questions PDF By ScholarHat
MVC Interview Questions PDF By ScholarHatMVC Interview Questions PDF By ScholarHat
MVC Interview Questions PDF By ScholarHat
Scholarhat
 
Introduction to Banking System in India.ppt
Introduction to Banking System in India.pptIntroduction to Banking System in India.ppt
Introduction to Banking System in India.ppt
Dr. S. Bulomine Regi
 
Open Source and AI - ByWater Closing Keynote Presentation.pdf
Open Source and AI - ByWater Closing Keynote Presentation.pdfOpen Source and AI - ByWater Closing Keynote Presentation.pdf
Open Source and AI - ByWater Closing Keynote Presentation.pdf
Jessica Zairo
 
V2-NLC-Certificate-of-Completion_Learner.docx
V2-NLC-Certificate-of-Completion_Learner.docxV2-NLC-Certificate-of-Completion_Learner.docx
V2-NLC-Certificate-of-Completion_Learner.docx
302491
 
MathematicsGrade7-Presentation-July-12024.pptx
MathematicsGrade7-Presentation-July-12024.pptxMathematicsGrade7-Presentation-July-12024.pptx
MathematicsGrade7-Presentation-July-12024.pptx
nolicaliso1
 
SQL Server Interview Questions PDF By ScholarHat
SQL Server Interview Questions PDF By ScholarHatSQL Server Interview Questions PDF By ScholarHat
SQL Server Interview Questions PDF By ScholarHat
Scholarhat
 
FIRST AID PRESENTATION ON INDUSTRIAL SAFETY by dr lal.ppt
FIRST AID PRESENTATION ON INDUSTRIAL SAFETY by dr lal.pptFIRST AID PRESENTATION ON INDUSTRIAL SAFETY by dr lal.ppt
FIRST AID PRESENTATION ON INDUSTRIAL SAFETY by dr lal.ppt
ashutoshklal29
 
React Interview Question PDF By ScholarHat
React Interview Question PDF By ScholarHatReact Interview Question PDF By ScholarHat
React Interview Question PDF By ScholarHat
Scholarhat
 
7. Post Harvest Entomology and their control.pptx
7. Post Harvest Entomology and their control.pptx7. Post Harvest Entomology and their control.pptx
7. Post Harvest Entomology and their control.pptx
UmeshTimilsina1
 
RDBMS Lecture Notes Unit4 chapter12 VIEW
RDBMS Lecture Notes Unit4 chapter12 VIEWRDBMS Lecture Notes Unit4 chapter12 VIEW
RDBMS Lecture Notes Unit4 chapter12 VIEW
Murugan Solaiyappan
 
Open and Critical Perspectives on AI in Education
Open and Critical Perspectives on AI in EducationOpen and Critical Perspectives on AI in Education
Open and Critical Perspectives on AI in Education
Robert Farrow
 
BỘ ĐỀ THI HỌC SINH GIỎI CÁC TỈNH MÔN TIẾNG ANH LỚP 9 NĂM HỌC 2023-2024 (CÓ FI...
BỘ ĐỀ THI HỌC SINH GIỎI CÁC TỈNH MÔN TIẾNG ANH LỚP 9 NĂM HỌC 2023-2024 (CÓ FI...BỘ ĐỀ THI HỌC SINH GIỎI CÁC TỈNH MÔN TIẾNG ANH LỚP 9 NĂM HỌC 2023-2024 (CÓ FI...
BỘ ĐỀ THI HỌC SINH GIỎI CÁC TỈNH MÔN TIẾNG ANH LỚP 9 NĂM HỌC 2023-2024 (CÓ FI...
Nguyen Thanh Tu Collection
 
FINAL MATATAG Science CG 2023 Grades 3-10.pdf
FINAL MATATAG Science CG 2023 Grades 3-10.pdfFINAL MATATAG Science CG 2023 Grades 3-10.pdf
FINAL MATATAG Science CG 2023 Grades 3-10.pdf
maritescanete2
 
QCE – Unpacking the syllabus Implications for Senior School practices and ass...
QCE – Unpacking the syllabus Implications for Senior School practices and ass...QCE – Unpacking the syllabus Implications for Senior School practices and ass...
QCE – Unpacking the syllabus Implications for Senior School practices and ass...
mansk2
 

Recently uploaded (20)

2 Post harvest Physiology of Horticulture produce.pptx
2 Post harvest Physiology of Horticulture  produce.pptx2 Post harvest Physiology of Horticulture  produce.pptx
2 Post harvest Physiology of Horticulture produce.pptx
 
Parkinson Disease & Anti-Parkinsonian Drugs.pptx
Parkinson Disease & Anti-Parkinsonian Drugs.pptxParkinson Disease & Anti-Parkinsonian Drugs.pptx
Parkinson Disease & Anti-Parkinsonian Drugs.pptx
 
MATATAG CURRICULUM sample lesson exemplar.docx
MATATAG CURRICULUM sample lesson exemplar.docxMATATAG CURRICULUM sample lesson exemplar.docx
MATATAG CURRICULUM sample lesson exemplar.docx
 
View Inheritance in Odoo 17 - Odoo 17 Slides
View Inheritance in Odoo 17 - Odoo 17  SlidesView Inheritance in Odoo 17 - Odoo 17  Slides
View Inheritance in Odoo 17 - Odoo 17 Slides
 
Benchmarking Sustainability: Neurosciences and AI Tech Research in Macau - Ke...
Benchmarking Sustainability: Neurosciences and AI Tech Research in Macau - Ke...Benchmarking Sustainability: Neurosciences and AI Tech Research in Macau - Ke...
Benchmarking Sustainability: Neurosciences and AI Tech Research in Macau - Ke...
 
Odoo 17 Events - Attendees List Scanning
Odoo 17 Events - Attendees List ScanningOdoo 17 Events - Attendees List Scanning
Odoo 17 Events - Attendees List Scanning
 
MVC Interview Questions PDF By ScholarHat
MVC Interview Questions PDF By ScholarHatMVC Interview Questions PDF By ScholarHat
MVC Interview Questions PDF By ScholarHat
 
Introduction to Banking System in India.ppt
Introduction to Banking System in India.pptIntroduction to Banking System in India.ppt
Introduction to Banking System in India.ppt
 
Open Source and AI - ByWater Closing Keynote Presentation.pdf
Open Source and AI - ByWater Closing Keynote Presentation.pdfOpen Source and AI - ByWater Closing Keynote Presentation.pdf
Open Source and AI - ByWater Closing Keynote Presentation.pdf
 
V2-NLC-Certificate-of-Completion_Learner.docx
V2-NLC-Certificate-of-Completion_Learner.docxV2-NLC-Certificate-of-Completion_Learner.docx
V2-NLC-Certificate-of-Completion_Learner.docx
 
MathematicsGrade7-Presentation-July-12024.pptx
MathematicsGrade7-Presentation-July-12024.pptxMathematicsGrade7-Presentation-July-12024.pptx
MathematicsGrade7-Presentation-July-12024.pptx
 
SQL Server Interview Questions PDF By ScholarHat
SQL Server Interview Questions PDF By ScholarHatSQL Server Interview Questions PDF By ScholarHat
SQL Server Interview Questions PDF By ScholarHat
 
FIRST AID PRESENTATION ON INDUSTRIAL SAFETY by dr lal.ppt
FIRST AID PRESENTATION ON INDUSTRIAL SAFETY by dr lal.pptFIRST AID PRESENTATION ON INDUSTRIAL SAFETY by dr lal.ppt
FIRST AID PRESENTATION ON INDUSTRIAL SAFETY by dr lal.ppt
 
React Interview Question PDF By ScholarHat
React Interview Question PDF By ScholarHatReact Interview Question PDF By ScholarHat
React Interview Question PDF By ScholarHat
 
7. Post Harvest Entomology and their control.pptx
7. Post Harvest Entomology and their control.pptx7. Post Harvest Entomology and their control.pptx
7. Post Harvest Entomology and their control.pptx
 
RDBMS Lecture Notes Unit4 chapter12 VIEW
RDBMS Lecture Notes Unit4 chapter12 VIEWRDBMS Lecture Notes Unit4 chapter12 VIEW
RDBMS Lecture Notes Unit4 chapter12 VIEW
 
Open and Critical Perspectives on AI in Education
Open and Critical Perspectives on AI in EducationOpen and Critical Perspectives on AI in Education
Open and Critical Perspectives on AI in Education
 
BỘ ĐỀ THI HỌC SINH GIỎI CÁC TỈNH MÔN TIẾNG ANH LỚP 9 NĂM HỌC 2023-2024 (CÓ FI...
BỘ ĐỀ THI HỌC SINH GIỎI CÁC TỈNH MÔN TIẾNG ANH LỚP 9 NĂM HỌC 2023-2024 (CÓ FI...BỘ ĐỀ THI HỌC SINH GIỎI CÁC TỈNH MÔN TIẾNG ANH LỚP 9 NĂM HỌC 2023-2024 (CÓ FI...
BỘ ĐỀ THI HỌC SINH GIỎI CÁC TỈNH MÔN TIẾNG ANH LỚP 9 NĂM HỌC 2023-2024 (CÓ FI...
 
FINAL MATATAG Science CG 2023 Grades 3-10.pdf
FINAL MATATAG Science CG 2023 Grades 3-10.pdfFINAL MATATAG Science CG 2023 Grades 3-10.pdf
FINAL MATATAG Science CG 2023 Grades 3-10.pdf
 
QCE – Unpacking the syllabus Implications for Senior School practices and ass...
QCE – Unpacking the syllabus Implications for Senior School practices and ass...QCE – Unpacking the syllabus Implications for Senior School practices and ass...
QCE – Unpacking the syllabus Implications for Senior School practices and ass...
 

Analyzing User Modeling on Twitter for Personalized News Recommendations

  • 1. Analyzing User Modeling on Twitter for Personalized News Recommendations Best Paper of Conference on User Modeling, Adaption and Personalization (UMAP’11) Authors: Fabian Abel, Qi Gao, Geert-Jan Houben and Ke Tao Unit for Social Software Presenter: Guangyuan Piao Reading Group, 29/09/2015
  • 2. Contents • Background & Related Work • Design Options for User Modeling in Online Social Networks for Recommendations • Research Questions • Dataset for the study • Study for Research Questions 1,2 • Experiment for Research Question 3 2/21
  • 4. User Modeling in Online Social Networks for Recommendations 4/21
  • 5. Related Work Representation of a User Model • bag-of-words, Chen et al. for recommending and ranking URLs posted in Twitter messages Study of hashtags in Twitter • investigate the specificity, stability over time, Laniado & Mika • temporal dynamics of hashtags, Laniado & Mika / Huang et al. Enrichment of tweets • Exploit metadata of research papers to enrich the semantics of tweets for mapping tweets to conference talks, Rowe et al. 5/21
  • 6. Design Options for User Modeling Semantic Enrichment Profile Type Time Constraint 1. tweet-based 2. further enrichment 1. hashtag-based 2. entity-based 3. topic-based 1. time period 2. temporal patterns User Profile Google … iPhone 0.09 … 0.08 P(u) = link news monitored news pool an entity-based profile 6/21
  • 7. Research Questions 1. how does the semantic enrichment impact the characteristics and quality of Twitter-based profiles? 2. how do (different types of) profiles evolve over time? Are there any characteristic patterns? 3. how do the different user modeling strategies impact personalization (personalized news recommendation) and does the consideration of temporal patterns improve the accuracy of the recommendations? 7/21
  • 8. Dataset for the study Collected from Twitter with more than • 20,000 Twitter users • 2 months • 10,000,000 tweets • 75,000 news articles Sample dataset for study • 1,619 users with at least 20 tweets & 1 tweet/month (2,316,204 tweets) 8/21
  • 9. Research Question 1 how does the semantic enrichment impact the characteristics and quality of Twitter-based profiles? 9/21
  • 10. • tweet-only-based user modeling, fails to create profiles for 100 users • by enrichment with entities and topics obtained from linked news articles ➡️ a higher # of distinct concepts and variety for per profile 10/21 The impact of news-based enrichment
  • 11. • entity-based and topic-based strategies have higher coverage Comparison of different types of profiles 11/21
  • 12. Research Question 2 how do (different types of) profiles evolve over time? are there any characteristic patterns? 12/21
  • 13. • profile difference is measured by d1-distance • user profiles change over time: the older a profile the more it differs from the current profile of the user Profile difference over time 13/21 d1 = 2 iPad iPhone Px(u) 1 0 Py(u) 0 1
  • 14. • the difference of weekday and weekend profiles is higher than that of other temporal patterns Profile difference of weekday and weekend 14/21
  • 15. • hashtag-based & entity-based profiles change most while the types of entities people refer to (person, product etc.) do not differ strongly Profile difference of weekday and weekend 15/21
  • 16. how do the different user modeling strategies impact personalization (personalized news recommendation)? and does the consideration of temporal patterns improve the accuracy of the recommendations? Research Question 3 16/21
  • 17. Evaluation Setup for News Recommendations Main goal: analyze and compare the applicability of the different user modeling strategies in the context of news recommendations. Recommendation algorithm: cosine similarity between user profile and tweets. Evaluation Metrics: • MRR (Mean Reciprocal Rank): at which rank the first item relevant to the user occurs on average. • S@k (Success at rank k): the mean probability that a relevant item occurs within the top k of the ranking, k=10. observation time period for Twitter last week used for user modeling get candidate news items 17/21
  • 18. Results for News Recommendations – impact of news-based enrichment • entity-based strategy is better than others • exploiting both tweets and linked news articles for creating user profiles improves the performance significantly (p < 0.05) 18/21
  • 19. Results for News Recommendations – impact of considering temporal patterns • fresh user profiles for topic-based user modeling are more applicable for recommending news articles, while complete user profiles for entity-based user modeling yields better recommendations. • similar patterns can be found for weekend news recommendations. (fresh profile: two weeks before recommendation time) 19/21
  • 20. Conclusions Further enrichment with semantics extracted from news articles • enhanced the variety of the constructed profiles • improved the accuracy of news article recommendations Temporal dynamics of user profiles • user profiles change over time • user vary from weekdays and weekends • consideration of temporal dynamics is beneficial to news recommendations for topic-based user modeling strategy 20/21

Editor's Notes

  1. Why not all?
  2. Ground tru
  3. Fresh user profile means?