RESUMO
Sentiment analysis (SA) aims to understand the attitudes and views of opinion holders with computers. Previous studies have achieved significant breakthroughs and extensive applications in the past decade, such as public opinion analysis and intelligent voice service. With the rapid development of deep learning, SA based on various modalities has become a research hotspot. However, only individual modality has been analyzed separately, lacking a systematic carding of comprehensive SA methods. Meanwhile, few surveys covering the topic of multimodal SA (MSA) have been explored yet. In this article, we first take the modality as the thread to design a novel framework of SA tasks to provide researchers with a comprehensive understanding of relevant advances in SA. Then, we introduce the general workflows and recent advances of single-modal in detail, discuss the similarities and differences of single-modal SA in data processing and modeling to guide MSA, and summarize the commonly used datasets to provide guidance on data and methods for researchers according to different task types. Next, a new taxonomy is proposed to fill the research gaps in MSA, which is divided into multimodal representation learning and multimodal data fusion. The similarities and differences between these two methods and the latest advances are described in detail, such as dynamic interaction between multimodalities, and the multimodal fusion technologies are further expanded. Moreover, we explore the advanced studies on multimodal alignment, chatbots, and Chat Generative Pre-trained Transformer (ChatGPT) in SA. Finally, we discuss the open research challenges of MSA and provide four potential aspects to improve future works, such as cross-modal contrastive learning and multimodal pretraining models.
RESUMO
Both traditional teaching and online teaching advocate individualized education. One of the difficulties on exploring possible improvements of instructional design is the challenging process of data collection. Existing research mainly focuses on the exam score of students but pays little attention to students' daily practice. As an effective method to handle time-series dataset, the generalized estimating equations (GEE) have not been used in this research field. Considering above issues, we first propose an experimental paradigm of programming performance analysis based on the performance record of students' daily practice-exam and finish collecting a complete time-series dataset in one semester, including students' individual attributes, learning behavior, and learning performance. Then, we propose an approach that analyzes practice-exam time-series dataset based on GEE to study the influence of individual attributes and learning behavior on learning performance. It is the first time to apply the GEE method for ordinal multinomial responses in this research field, by which we conclude several results that gender or major does have a certain difference on the programming learning. The longer the answer time and the less the cost time, the better the students' performance. Regardless of gender, students tend to cram for the exam and perform a little worse in the daily exercise. Finally, targeting at two important individual attributes, we give corresponding teaching mode decisions that university should teach students programming by major and teacher should give different teaching methods to students of different genders at different time points.