exploiting social relations for sentiment analysis in...

22
Data Mining and Machine Learning Lab Exploiting Social Relations for Sentiment Analysis in Microblogging Xia Hu 1 , Lei Tang 2 , Jiliang Tang 1 , Huan Liu 1 1 Arizona State University 2 @WalmartLabs ACM International Conference on Web Search and Data Mining (WSDM 2013)

Upload: others

Post on 17-Jul-2020

9 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Data Mining and Machine Learning Lab

Exploiting Social Relations for Sentiment Analysis

in Microblogging

Xia Hu1, Lei Tang2, Jiliang Tang1, Huan Liu1 1 Arizona State University

2@WalmartLabs

ACM International Conference on Web Search and Data Mining (WSDM 2013)

Page 2: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Outline

• Motivation

• Sentiment Analysis with Social Relations

• Experimental Evaluation

• Conclusions and Future Work

Page 3: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Why Sentiment Analysis in Microblogging?

Microblogging has become a popular form of social media,

through which users can easily generate comments on

breaking news, public events, or products.

Page 4: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Traditional Data

Pure Textual Content

Page 5: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Microblogging Data

Page 6: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Microblogging Data

Page 7: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Social Theories

Sentiment Consistency suggests that the sentiments of two messages posted

by the same user are more likely to be consistent than

those of two random messages.

Emotional Contagion reveals that the sentiments of two messages

posted by friends are more likely to be similar

than those of two random messages.

Page 8: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Hypothesis Testing (I)

Datasets

Page 9: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Hypothesis Testing (II)

Sentiment Difference Score

Verifying Sentiment Consistency

Page 10: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Hypothesis Testing (III)

Sentiment Difference Score

Verifying Sentiment Contagion

Page 11: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Outline

• Motivation

• Sentiment Analysis with Social Relations

• Experimental Evaluation

• Conclusions and Future Work

Page 12: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Sentiment Analysis

= Coefficients × FeatureVector(TWEET)

Sentiment (TWEET )

Page 13: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Modeling Social Relations

Page 14: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Sentiment Analysis with Social Relations

= Coefficients × FeatureVector(TWEET)

Sentiment (TWEET )

Textual

Information

Social

Relations

Page 15: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

A Sparse Formulation

= Coefficients × FeatureVector(TWEET)

Sentiment (TWEET)

Textual

Information

Social

Relations

Sparse

Regularization

Page 16: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Outline

• Motivation

• Sentiment Analysis with Social Relations

• Experimental Evaluation

• Conclusions and Future Work

Page 17: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Experiments (I)

Comparison with Text-based Methods

Page 18: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Experiments (II)

Incorporating Social Relations

Page 19: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Outline

• Motivation

• Sentiment Analysis with Social Relations

• Experimental Evaluation

• Conclusions and Future Work

Page 20: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Conclusions

•We formally define the problem of sentiment analysis with

social relations in microblogging

•By verifying the existence of two social theories in

microblogging, we build sentiment relations between

messages via social relations

•We present a novel supervised method to tackle the

high-dimensional texts by integrating sentiment relations

between the texts

Page 21: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Future Work

•Contextual information, like spatial-temporal patterns, could

be potentially useful to measure the sentiment consistency of

people as well

•We can further explore how sentiments diffuse in the social

network and how people's sentiments correlate with internal

(their friends) and external (public events) factors

Page 22: Exploiting Social Relations for Sentiment Analysis in ...faculty.cs.tamu.edu/xiahu/slides/WSDM2013_slides.pdfWhy Sentiment Analysis in Microblogging? Microblogging has become a popular

Questions

Acknowledgments: This work is, in part, sponsored by ONR (N000141110527) and

(N000141010091). Comments and suggestions from DMML members and reviewers are

greatly appreciated.