Αυξητική μέθοδος ανάλυσης συναισθήματος σε περιβάλλον microblogging
View/
Keywords
Short texts ; Incremental Learning ; Μηχανική μάθηση ; Μικρά κείμενα ; Machine learning ; Sentiment analysis ; Ανάλυση συναισθήματος ; Μάθηση με αυξητικό τρόποAbstract
Machine learning tools that perform sentiment analysis on texts is one of the most innovative branches of machine learning. To train such tools, large datasets are usually required. In some cases, these datasets are fully available in their entirety, but in other, more realistic cases, the data can be accessed incrementally, as they are created. In order to utilize such datasets as well as perform analyses on real data without having access to the whole dataset from the beginning, machine learning tools that can be trained incrementally are used. The focus of this thesis is this novel and innovative branch of sentiment analysis machine learning tools that can be trained incrementally.
This master thesis aims to find the optimal classifier, which can be trained incrementally, for classifying short texts based on their sentiment. The thesis is divided into four chapters. In Chapter 1 the most common methods to represent text into vectors are discussed as well as the one chosen for this analysis. In Chapter 2 various classifiers that can be trained incrementally are detailed. Five criteria are devised, and the best classifier is identified. In Chapter 3 a detailed analysis of the optimal classifier is conducted. That classifier is put through various tests and its behavior is studied. Finally, an integrated tool based on the chosen classifier, is created in order to classify short texts based in their sentiment. Chapter 4 is the conclusion of this thesis.
Postgraduate Studies Programme
Ψηφιακά Συστήματα και ΥπηρεσίεςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών ΣυστημάτωνNumber of pages
103Language
GreekCollections
Except where otherwise noted, this item's license is described as
Attribution-NonCommercial-NoDerivatives 4.0 Διεθνές
Attribution-NonCommercial-NoDerivatives 4.0 Διεθνές