The MuSe 2021 Multimodal Sentiment Analysis Challenge

Stappen, Lukas; Baird, Alice; Christ, Lukas; Schumann, Lea; Sertolli, Benjamin; Meßner, Eva-Maria; Cambria, Erik; Zhao, Guoying; Schuller, Björn W.

doi:10.1145/3475957.3484450

Cited by 62 publications

(13 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the MuSe-Personalisation Sub-Challenge, participants are encouraged to develop methods for tailoring models to specific individuals. MuSe-Personalisation employs the Ulm-TSST dataset already featured in previous iterations of MuSe [19,54]. While all sub-challenges feature rather simple scenarios involving one person, we believe that the variety of prediction targets and modalities will also foster progress in empathetic dialogue systems and conversational sentiment analysis.…”

Section: Discussionmentioning

confidence: 99%

“…The training dataset comprises 41, both development and test sets 14 videos. Note that the split is identical to the splits of Ulm-TSST employed for 2021's and 2022's Emotional Stress Sub-Challenge (MuSe-Stress) challenges [19,54]. In order to facilitate personalisation on the test subjects, we provide labelled parts of their videos as follows: we take the first 60 seconds of each test video and consider this our subject-specific training data.…”

Section: The Muse-personalisation Sub-challengementioning

confidence: 99%

“…Window size and step length of the segmentation are optimised within the hyperparameter search. Analogously to previous MuSe challenges [19,54,55], we employ CCC-Loss as the loss function. We limit our experiments to audio and video-based features and leave experimentation with both physiological signals and textual features to the participants.…”

Section: Muse-mimicmentioning

confidence: 99%

“…Thus, MuSe-Personalisation encourages the exploration of the adaptation of multimodal emotion recognition systems to individuals, taking their specific features into account. MuSe-Personalisation utilises the Ulm-Trier Social Stress Test dataset (Ulm-TSST) introduced in MuSe 2021 [54,56]. Ulm-TSST consists of recordings of 69 individuals undergoing the Trier Social Stress Test (TSST), a scenario designed to induce stress.…”

mentioning

confidence: 99%

See 3 more Smart Citations

The MuSe 2022 Multimodal Sentiment Analysis Challenge

Christ

Amiriparian

Baird³

et al. 2022

Proceedings of the 3rd International on Multimodal Sentiment Analysis Workshop and Challenge

Self Cite

View full text Add to dashboard Cite

nnnnnnn feature sets. As a competitive baseline system, a Gated Recurrent Unit (GRU)-Recurrent Neural Network (RNN) is employed. On the respective sub-challenges' test datasets, it achieves a mean (across three continuous intensity targets) Pearson's Correlation Coefficient of .4727 for MuSe-Mimic, an Area Under the Curve (AUC) value of .8310 for MuSe-Humour and Concordance Correlation Coefficient (CCC) values of .7482 for arousal and .7827 for valence in the MuSe-Personalisation sub-challenge. CCS CONCEPTS• Computing methodologies → Neural networks; Artificial intelligence; Computer vision; Natural language processing.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: The Muse-personalisation Sub-challengementioning

confidence: 99%

Section: Muse-mimicmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

The MuSe 2022 Multimodal Sentiment Analysis Challenge

Christ

Amiriparian

Baird³

et al. 2022

Proceedings of the 3rd International on Multimodal Sentiment Analysis Workshop and Challenge

Self Cite

View full text Add to dashboard Cite

show abstract

“…Potential customers' purchasing decisions are influenced by online analyses of e-commerce, including userinitiated remarks about caliber of products, services, [18]etc., which also directly affect how sickly users adhere to e-commerce platforms. Sentiment analysis(SA) is the act of sifting through these evaluations to find favorable and unfavorable opinions in order to determine a person's likelihood of purchasing a product [19].…”

Section: Introductionmentioning

confidence: 99%

Emotion Analysis of Literary Works Based on Qutrit-inspired Fully Self-supervised Quantum Neural Network

Lirong Wang

2024

jes

View full text Add to dashboard Cite

The development of foreign literature, embodiment of emotional value in modern, contemporary foreign literature is more attentive on experience, through reading it, one can understand humanistic, personal feelings embedded in the work and grasp the author’s personality, spiritual experience. To analyze sentiment data of foreign literary works, proposed Emotion Analysis of Literary Works Based on Qutrit-inspired Fully Self-supervised Quantum Neural Network Method (EA-LW- QIFSQN). Initially input data are collected from NLP-dataset. Afterward, the input data provided to preprocessing. In preprocessing segment Federated Neural Collaborative Filtering (FNCF) is used to clean the unwanted data. Then preprocessed data is fed to feature extraction, synchro spline-kernelled chirplet extracting transform (SKCET) is used to extract two features such as textual features and lexical features. Afterwards QIFSQN is used to classify the emotions likes joy, sadness, anger, fear. Generally, QIFSQN doesn’t show some optimization adaption techniques to determine optimum parameter to offer accurate detection. Polar Coordinate Bald Eagle Search Algorithm (PCBSOA) is proposed to enhance QICCN classifies the emotions accurately. The proposed technique is executed and efficacy of EA-LW- QIFSQN technique is assessed with support of numerous performances like accuracy, recall, precision and F1-scorce is analyzed. Then, performance of EA-LW- QIFSQN technique is analyzed with existing techniques like emotion analysis of literary works depend on attention mechanisms with fusion of two-channel features (EA-LW-CNN), integrative improvement of modern literary works with traditional culture combined by semantic association network modeling(EA-LW-SAN) and emotion expression in modern literary appreciation: emotion-depend analysis (EA-LW-RFA) respectively.

show abstract