Extraction Method for Important Words as a Viewer’s Reaction Arousal Factor from YouTube - Transcription

Abstract

We present a novel extraction method for important words as a viewer’s reaction arousal factor from YouTube transcription. The proposed method analyses the content of social media posts. Further, it extracts important words that are likely or unlikely to evoke reactions from the viewers. In this study, we analyze the subtitles and the statistical data obtained from YouTube videos. Further, a database is created consisting of the extracted words that are likely or unlikely to evoke reactions. The method consists of obtaining the subtitle data and the statistical data from YouTube, building a database, and constructing a machine learning model to classify them. This is followed by the local interpretation of the model to extract the aforementioned words. The experimental results showed that the machine learning model was effective using the created database.