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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Multimodal sentiment analysis is a core research area that studies speakersentiment expressed from the language, visual, and acoustic modalities. Thecentral challenge in multimodal learning involves inferring jointrepresentations that can process and relate information from these modalities.However, existing work learns joint representations by requiring all modalitiesas input and as a result, the learned representations may be sensitive to noisyor missing modalities at test time. With the recent success of sequence tosequence (Seq2Seq) models in machine translation, there is an opportunity toexplore new ways of learning joint representations that may not require allinput modalities at test time. In this paper, we propose a method to learnrobust joint representations by translating between modalities. Our method isbased on the key insight that translation from a source to a target modalityprovides a method of learning joint representations using only the sourcemodality as input. We augment modality translations with a cycle consistencyloss to ensure that our joint representations retain maximal information fromall modalities. Once our translation model is trained with paired multimodaldata, we only need data from the source modality at test time for finalsentiment prediction. This ensures that our model remains robust fromperturbations or missing information in the other modalities. We train ourmodel with a coupled translation-prediction objective and it achieves newstate-of-the-art results on multimodal sentiment analysis datasets: CMU-MOSI,ICT-MMMO, and YouTube. Additional experiments show that our model learnsincreasingly discriminative joint representations with more input modalitieswhile maintaining robustness to missing or perturbed modalities.

Found in Translation: Learning Robust Joint Representations by Cyclic Translations Between Modalities