MCPR: A Chinese Product Review Dataset for Multimodal Aspect-Based Sentiment Analysis

Aspect-based sentiment analysis (ABSA), which aims to analyze the sentiments toward the extracted aspects, has been attracting considerable interest in the last decade. Most of the existing studies concentrate on determining the sentiment polarity of the given aspect according to only textual content, while there is little research on multimodal aspect-based sentiment analysis (MABSA) due to the scarcity of datasets consisting of multimodality content, such as both texts and images. In this paper, we design and construct a Multimodal Chinese Product Review dataset (MCPR) to support the research of MABSA. MCPR is a collection of 1.5k product reviews involving clothing and furniture departments, from the e-commercial platform JD.com. After aspect-base sentiment annotation and text-image matching, we obtain 2,719 text-image pairs and 610 distinct aspects in total. It is the first aspect-based multimodal Chinese product review dataset.