Review of sentiment analysis in new product development: text, audio, visual, and multimodal data

Abstract This research provides a literature review on the application of sentiment analysis (SA) in the new product development (NPD) process. The literature review employs a systematic literature review methodology. The steps include selecting a review topic, searching and selecting relevant articles, assessing and synthesising the literature, and organising the writing of the review. Sentiment analysis is a subdomain of natural language processing (NLP) that examines user opinions. The sentiment analysis methodology has been employed in the new product development process to replace traditional methods. Sentiment analysis can be conducted across various modalities, including text, audio, image, and multimodal formats. Text modality for sentiment analysis has been used to enhance the lifetime of products and services. Audio data and image modalities represent alternative modalities; however, they receive significantly less attention. The limitation is that these modalities are predominantly executed in controlled environments, utilising open-source or benchmark datasets, and some still employ text modality sentiment analysis methods or lexicons. Multimodal data, conversely, aims to augment the informational dimension of the text modality and is typically executed using deep learning models. This modality encompasses numerous combinations with the primary objective of enhancing the performance of sentiment analysis, hence reducing bias. The findings suggest that future research in this domain should focus on improving multimodal sentiment analysis to improve the new product development process.