Sentiment Analysis about Turkish TV Series with Web Scraping

Abstract

A person's preferences for a product or service are influenced by other people's opinions about that product or service. One of the sectors where this situation is most common is digital media platforms that broadcast TV series. Therefore, the automatic extraction of other people's feelings and opinions from comments about TV series on digital media platforms by sentiment analysis allows users to have information and judgment about series easily. This study aims to conduct sentiment analysis on the Turkish Tv Series with machine learning algorithms. First, we collect comments about digital media platform TV series from a user-generated website, namely Ekşi Sözlük, with web scraping techniques and apply preprocessing to this data. Then, we conduct sentiment analysis with Support Vector Machine (SVM), Logistic Regression (LR), K-nearest Neighbors machine learning techniques with Bag of Words (BoW), Term Frequency Inverse Document Frequency (TF-IDF), Word2vec vector models. Our results show that the SVM classifier trained on the TF -IDF vector model gives the highest prediction macro averaged F -score with 0.631.