A Corpus for Dimensional Sentiment Classification on YouTube Streaming Service

Abstract The streaming service platform such as YouTube provides a discussion function for audiences worldwide to share comments. YouTubers who upload videos to the YouTube platform want to track the performance of these uploaded videos. However, the present analysis functions of YouTube only provide a few performance indicators such as average view duration, browsing history, variance in audience’s demographics, etc., and lack of sentiment analysis on the audience's comments. Therefore, the paper proposes multi-dimensional sentiment indicators such as YouTuber preference, Video preferences, and Excitement level to capture comprehensive sentiment on audience comments for videos and YouTubers. To evaluate the performance of different classifiers, we experiment with deep learning- based, machine learning-based, and BERT-based classifiers to automatically detect three sentiment indicators of an audience's comments. Experimental results indicate that the BERT-based classifier is a better classification model than other classifiers according to F1-score, and the sentiment indicator of Excitement level is quite an improvement. Therefore, the multiple sentiment detection tasks on the video streaming service platform can be solved by the proposed multi- dimensional sentiment indicators accompanied with BERT classifier to gain the best result. Keywords: Sentiment Analysis, Text Classification, Machine Learning, Deep Learning, Streaming Service 1 Introduction Due to the rapid rise of new media and the popularization of mobile phone networks, audiences’ viewing habits have shifted from TV to online social media platforms. Now people can watch videos on different platforms such as Facebook, Dailymotion, and YouTube anytime and anywhere. YouTube has 16 million active users in Taiwan monthly, and nearly 93% of users have visited YouTube. In addition, YouTube has become ubiquitous and played an increasingly important role in modern life and entertainment. Also, YouTube provides a discussion function for audiences to express their opinion by clicking like or dislike bottom or leaving comments. Therefore, comprehensive sentiment analysis for comments of the audience on YouTube is necessary. It is verified that public views, comments, and attitudes towards many events can be analyzed through social media (Heredia et al., 2016). Public reviews on Amazon were used to evaluate users’ opinions and determine the audience’s preference by classifying opinions into negative, positive, and neutral (Bhatt et al., 2015). Another research investigated the popularity of videos by indicators such as the number of likes, dislikes, and views (Chelaru et al., 2013). Social media, especially YouTube. is considered the largest video sharing site, and the platform has developed into a leading marketing tool. (Schwemmer and Ziewiecki, 2018) Inspired by the above analysis tasks and the rapid growth status of YouTube, we propose the multi- dimensional sentiment indicators to analyze comments on YouTube, which aim to help YouTubers check their videos’ performance uploaded on the YouTube platform. In general, sentiment analysis focuses on determining the positive, negative, or neutral emotions in many pieces of research (Cunha et al., 2019). Even if Keith et al. (2016) extend the emotional detection, which includes highly positive, optimistic, neutral, negative, and highly damaging, the emotional variance may have a different dimension, such as excitement which expresses the …

Content truncated to 500 words