Συναισθηματική ανάλυση δημοσιεύσεων χρηστών κοινωνικών δικτύων με εφαρμογή στο Twitter
Sentiment analysis on Twitter Greek data
View/
Keywords
NLP ; Sentiment analysis ; Επεξεργασία φυσικής γλώσσας ; Κοινωνικά δίκτυα ; Machine learning ; Python ; Twitter ; Ανάλυση συναισθήματος ; Μηχανική μάθηση ; Social mediaAbstract
Sentiment analysis or opinion mining is a very interesting scientific field that is currently gaining a lot of interest in a lot of different fields. By using Sentiment Analysis techniques, we can predict efficiently people’s mind and behavior around a specific field. Social Media nowadays are not only a way to communicate but they have become a powerful way to manipulate as well as influence individuals. Social media have become a big part of individual’s everyday life. A big percentage of nowadays data is located on social media. Specifically, Twitter is a very powerful tool with incredible scientific interest in sentiment analysis field because of the incredible amount of data that a researcher can mine easily through Twitter API. This research is proposing a Greek sentiment analysis system from scratch. Our dataset consists of tweets collected from Twitter Streaming API in Greek language. The topic that those data are about is the Covid-19 vaccine that has been a very popular topic some months now.
This thesis is proposing a hybrid approach of sentiment analysis that can be used in Greek data. We first collected the dataset that consists of tweets in Greek language. Secondly, we used a rule-based method for classify the tweets in the right sentiment class. Thirdly, we trained a machine learning classifier that used the rated tweets as an input in order to predict the sentiment of a new dataset that we collected. Finally, we made a sentiment analysis simple app by deploying the machine learning model using Flask.