An Approach to Subjectivity Detection on Twitter Using the Structured Information
0 likes95 views
The document proposes using structured information from Twitter, like hashtags and retweets, to improve sentiment analysis. It describes related work using social and linguistic features for sentiment analysis. An experiment combines text and structured attributes from tweets to train classifiers, finding a stacking model using both achieved the best performance. Future work could explore more Twitter elements and their relation to sentiment.
![Improving the Sentiment Analysis ...
Related Work
► Contextual Applications in Sentiment Analysis
► [Pennacchiotti and Popescu, 2011] Linguistic and social network.
► [De Choudhury et al., 2013] User behavior to predict emotional states.
► Classification algorithms
► [Cortes and Vapnik, 1995] Support Vector Machine (SVM)
► [Cox, 1958] Logistic Regression (LR)
► [Friedman, 2001] Gradient Boosting Classifier (GBC)
► Train and Test Dataset
► [Villena-Román et al., 2015] TASS’15 General Corpus.
► 7.219 (11%) Train / 60.798 (89%) Test.
► Six different polarity labels: P+, P, N+, N, NEU, NONE
Related Work 55/15](https://image.slidesharecdn.com/iccci16-160926153545/85/An-Approach-to-Subjectivity-Detection-on-Twitter-Using-the-Structured-Information-5-320.jpg)
![Improving the Sentiment Analysis ...
Experiments
Experiments 1010/15
► Selected features to train a classifier
► [Barbosa and Feng, 2010]
► URL
► Exclamation marks
► Emoticons
► Uppercase words
► Uppercase Percent
► Favorites
► Modularity Class
► Directed graph relations based on “Follow”
► Three communities formed by left/right/neutral ideologies.
► Graph Degrees (In-Degree - Out-Degree)
► Retweets (RTs)
► Ellipsis](https://image.slidesharecdn.com/iccci16-160926153545/85/An-Approach-to-Subjectivity-Detection-on-Twitter-Using-the-Structured-Information-10-320.jpg)
An Approach to Subjectivity Detection on Twitter Using the Structured Information
- 1. Improving the Sentiment Analysis ... DeustoTech - Deusto Institute of Technology, University of Deusto http://www.morelab.deusto.es September 28, 2016 An Approach to Subjectivity Detection on Twitter Using the Structured Information ICCCI 2016 - 8th International Conference on Computational Collective Intelligence Juan Sixto, Aitor Almeida and Diego López-de-Ipiña 1
- 2. Improving the Sentiment Analysis ... Overview Introduction & Motivation Related Work Sentiment Analysis of Twitter Data Structured and Unstructured Information Experiments Conclusions & Future Work 2/15
- 3. Improving the Sentiment Analysis ... Introduction & Motivation ► User-generated information of social networks ► New algorithms and methods for their classification. ► The Sentiment Analysis (SA) methods. ► Ranking algorithms as resources. ► Microblogging and Twitter ► One of the largest textual data sources. ► Specific characteristics. Introduction & Motivation 33/15
- 4. Improving the Sentiment Analysis ... Introduction & Motivation ► Can the Structured Information of Twitter be used for sentiment analysis at global level? ► How the Structured Information of Twitter is classified? ► What Structural features are useful to subjectivity detection task? Introduction & Motivation 44/15
- 5. Improving the Sentiment Analysis ... Related Work ► Contextual Applications in Sentiment Analysis ► [Pennacchiotti and Popescu, 2011] Linguistic and social network. ► [De Choudhury et al., 2013] User behavior to predict emotional states. ► Classification algorithms ► [Cortes and Vapnik, 1995] Support Vector Machine (SVM) ► [Cox, 1958] Logistic Regression (LR) ► [Friedman, 2001] Gradient Boosting Classifier (GBC) ► Train and Test Dataset ► [Villena-Román et al., 2015] TASS’15 General Corpus. ► 7.219 (11%) Train / 60.798 (89%) Test. ► Six different polarity labels: P+, P, N+, N, NEU, NONE Related Work 55/15
- 6. Improving the Sentiment Analysis ... Sentiment Analysis of Twitter Data Okapi BM25 ranking function 66/15 ► Sentiment Analysis (or Opinion Mining) is defined as the task of finding the opinions of authors about specific entities. ► Feldman, 2013 ► Twitter text corpora ► Heterogeneous user-generated corpora ► Open Domain ► Noisy Text
- 7. Improving the Sentiment Analysis ... Structured and Unstructured Information Okapi BM25 ranking function 77/15
- 8. Improving the Sentiment Analysis ... Adaptation of the algorithm ► Four categories of attributes. ► Text attributes ► Hashtags, Links, Emoticons, Punctuation, Retweet,... ► Tweet attributes ► Quantity of retweets, creation date/time, associated place,... ► User attributes ► Location, political affiliation, post habits,... ► Topographic attributes ► Modularity class of user, In-degree, Out-degree, Communities,... Okapi BM25 ranking function 88/15
- 9. Improving the Sentiment Analysis ... Adaptation of the algorithm ► Four categories of attributes. ► Text attributes ► Hashtags, Links, Emoticons, Punctuation, Retweet,... ► Tweet attributes ► Quantity of retweets, creation date/time, associated place,... ► User attributes ► Location, political affiliation, post habits,... ► Topographic attributes ► Modularity class of user, In-degree, Out-degree, Communities,... Okapi BM25 ranking function 99/15
- 10. Improving the Sentiment Analysis ... Experiments Experiments 1010/15 ► Selected features to train a classifier ► [Barbosa and Feng, 2010] ► URL ► Exclamation marks ► Emoticons ► Uppercase words ► Uppercase Percent ► Favorites ► Modularity Class ► Directed graph relations based on “Follow” ► Three communities formed by left/right/neutral ideologies. ► Graph Degrees (In-Degree - Out-Degree) ► Retweets (RTs) ► Ellipsis
- 11. Improving the Sentiment Analysis ... Experiments ● Meta-Information classifier ○ GradientBoosting model ● Bag-of-Words classifier ○ Logistic Regression model ● Meta-Information and Bag-of-Words classifier ○ Matrix representation of structural features ○ GradientBoosting model ● Meta-Information and Bag-of-Words Stacking Classifier ○ Both models. ○ Array of level-0 models. ○ Logistic Regression model Experiments 1111/15
- 12. Improving the Sentiment Analysis ... Experiments ► Test datasets : 60.798 items. ► 6 categories: NONE,NEU,P,N,P+,N+ ► NONE: 20.54 % (Train) and 12,30 % (Test). ► Performance measures: ► Accuracy: true results / total dataset. ► Macro averaged-F1: precision and recall. ► NONE-F1: micro averaged F1 of the True labels. Experiments 1212/15
- 13. Improving the Sentiment Analysis ... Conclusions Conclusions 1313/15 ► We have proposed a method which: ► Adapt the contextual data to the global polarity detection task. ► Add new ways to use the contextual information. ► We presented a contextual data classification. ► We combined the structured and unstructured information to complement the classification task.
- 14. Improving the Sentiment Analysis ... Future Work Future Work 1414/15 ► Improve the present system including: ► More Twitter components and their relation with polarity. ► Lexicons and semantic resources. ► Extend the classifier to a global polarity task ► Study the relation between structural data and other user features.
- 15. Improving the Sentiment Analysis ... Thank You 1515/15
- 16. Improving the Sentiment Analysis ... All rights of images are reserved by the original owners*, the rest of the content is licensed under a Creative Commons by-sa 3.0 license.
- 17. Improving the Sentiment Analysis ... DeustoTech - Deusto Institute of Technology, University of Deusto http://www.morelab.deusto.es An Approach to Subjectivity Detection on Twitter Using the Structured Information Juan Sixto, Aitor Almeida and Diego López-de-Ipiña {jsixto, aitor.almeida, dipina }@deusto.es