dc.contributor.author |
Kamareddine, Mohamad Yahya, |
dc.date.accessioned |
2017-12-11T16:30:49Z |
dc.date.available |
2017-12-11T16:30:49Z |
dc.date.issued |
2017 |
dc.date.submitted |
2017 |
dc.identifier.other |
b19184256 |
dc.identifier.uri |
http://hdl.handle.net/10938/20971 |
dc.description |
Thesis. M.S. American University of Beirut. Computational Science Program, 2017. T:6599 |
dc.description |
Advisor : Dr. Mariette Awad, Associate Professor, Electrical and Computer Engineering ; Committee members : Dr. Nabil Nassif, Professor, Mathematics ; Dr. Mazen Al-Ghoul, Professor, Chemistry. |
dc.description |
Includes bibliographical references (leaves 130-133) |
dc.description.abstract |
With the richness of data and information, particularly for sports in social media, especially Twitter and Facebook, natural language processing could identify subjectivity and objectivity of phrases in one hand, and could extend analysis to pinpoint to the existence of consensus for these phrases on the other hand. However, finding the ground truth of a phrase or a sentence being subjective or objective, is complicated due to the underlying English context for some sentences that might embed different forms of subjectivity. Indeed, objectivity and subjectivity in phrases could be directly or indirectly established in a sentence. Moreover, humans could agree or disagree on different topics, in which each individual could express his-her opinions in a manner that might be with or against the considered topic. Motivated to apply artificial intelligence to sports phrases classification, this work presents an innovative sports related Twitter data analytics framework wrapped in a graphical user interface (GUI). This framework classifies sport phrases as subjective or objective, taking an additional step to introduce a new consensus label for subjective phrases according to Twitter bloggers. Our proposed workflow preprocesses and analyzes sport phrases, before generating decision trees based on tweets to identify objective and consensus-subjective polarities. Experimental results on a homemade corpus of 1007 phrases, reached an accuracy of 88percent on the test set for objectivity and subjectivity of phrases using our proposed set of syntactic and semantic features. The syntactic and semantic features were tested on five classifiers (KNN, SVM, NN, NB, and AB) and results were validated using various statistical measures. Moreover, applying the decision tree on the corpus recorded that 57percent of the subjective phrases are with consensus, while 43percent of them are without consensus. Furthermore, our framework extends analysis to identify that herding behaviors of bloggers lead to consensus on subjective phrases rather than object |
dc.format.extent |
1 online resource (xiii, 133 leaves) : illustrations |
dc.language.iso |
eng |
dc.relation.ispartof |
Theses, Dissertations, and Projects |
dc.subject.classification |
T:006599 |
dc.subject.lcsh |
Twitter. |
dc.subject.lcsh |
Data mining. |
dc.subject.lcsh |
Information science. |
dc.subject.lcsh |
Sports. |
dc.subject.lcsh |
Social media. |
dc.title |
An affective data science approach for sports related tweets - |
dc.type |
Thesis |
dc.contributor.department |
Faculty of Arts and Sciences. |
dc.contributor.department |
Computational Science Program, |
dc.contributor.institution |
American University of Beirut. |