Sentiment Analysis of a Text Story Dataset Collected Using Illustration Cards

초록

This paper introduces a crowdsourced text story dataset using the story card images of a commercial storytelling board game. We collect 705 text stories using illustration cards, where each story is classified with its ending type - either a happy or sad ending. Then, the sentiment patterns for a series of illustration cards matching each text story are analyzed to explore possible emotional arcs present in each story. The most dominant sentiment pattern is Negative-Negative-Positive (N-N-P) in happy-ending stories; Negative-Negative-Negative (N-N-N) in sad-ending stories. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.

키워드