English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Multimodal sentiment analysis has attracted increasing attention and lots ofmodels have been proposed. However, the performance of the state-of-the-artmodels decreases sharply when they are deployed in the real world. We find thatthe main reason is that real-world applications can only access the textoutputs by the automatic speech recognition (ASR) models, which may be witherrors because of the limitation of model capacity. Through further analysis ofthe ASR outputs, we find that in some cases the sentiment words, the keysentiment elements in the textual modality, are recognized as other words,which makes the sentiment of the text change and hurts the performance ofmultimodal sentiment models directly. To address this problem, we propose thesentiment word aware multimodal refinement model (SWRM), which can dynamicallyrefine the erroneous sentiment words by leveraging multimodal sentiment clues.Specifically, we first use the sentiment word position detection module toobtain the most possible position of the sentiment word in the text and thenutilize the multimodal sentiment word refinement module to dynamically refinethe sentiment word embeddings. The refined embeddings are taken as the textualinputs of the multimodal feature fusion module to predict the sentiment labels.We conduct extensive experiments on the real-world datasets includingMOSI-Speechbrain, MOSI-IBM, and MOSI-iFlytek and the results demonstrate theeffectiveness of our model, which surpasses the current state-of-the-art modelson three datasets. Furthermore, our approach can be adapted for othermultimodal feature fusion models easily. Data and code are available athttps://github.com/albertwy/SWRM.

Sentiment Word Aware Multimodal Refinement for Multimodal Sentiment Analysis with ASR Errors