In this work, a kernel attention module is presented for the task of EEG-based emotion classification with neural networks. The proposed module utilizes a self-attention mechanism by performing a kernel trick, demanding significantly fewer trainable parameters and computations than standard attention modules. The design also provides a scalar for quantitatively examining the amount of attention assigned during deep feature refinement, hence help better interpret a trained model. Using EEGNet as the backbone model, extensive experiments are conducted on the SEED dataset to assess the module's performance on within-subject classification tasks compared to other SOTA attention modules. Requiring only one extra parameter, the inserted module is shown to boost the base model's mean prediction accuracy up to more than 1\% across 15 subjects. A key component of the method is the interpretability of solutions, which is addressed using several different techniques, and is included throughout as part of the dependency analysis.
In this work, a parameter-efficient attention module is presented for emotion classification using a limited, or relatively small, number of electroencephalogram (EEG) signals. This module is called the Monotonicity Constrained Attention Module (MCAM) due to its capability of incorporating priors on the monotonicity when converting features' Gram matrices into attention matrices for better feature refinement. Our experiments have shown that MCAM's effectiveness is comparable to state-of-the-art attention modules in boosting the backbone network's performance in prediction while requiring less parameters. Several accompanying sensitivity analyses on trained models' prediction concerning different attacks are also performed. These attacks include various frequency domain filtering levels and gradually morphing between samples associated with multiple labels. Our results can help better understand different modules' behaviour in prediction and can provide guidance in applications where data is limited and are with noises.
Understanding human emotions is a crucial ability for intelligent robots to provide better human-robot interactions. The existing works are limited to trimmed video-level emotion classification, failing to locate the temporal window corresponding to the emotion. In this paper, we introduce a new task, named Temporal Emotion Localization in videos~(TEL), which aims to detect human emotions and localize their corresponding temporal boundaries in untrimmed videos with aligned subtitles. TEL presents three unique challenges compared to temporal action localization: 1) The emotions have extremely varied temporal dynamics; 2) The emotion cues are embedded in both appearances and complex plots; 3) The fine-grained temporal annotations are complicated and labor-intensive. To address the first two challenges, we propose a novel dilated context integrated network with a coarse-fine two-stream architecture. The coarse stream captures varied temporal dynamics by modeling multi-granularity temporal contexts. The fine stream achieves complex plots understanding by reasoning the dependency between the multi-granularity temporal contexts from the coarse stream and adaptively integrates them into fine-grained video segment features. To address the third challenge, we introduce a cross-modal consensus learning paradigm, which leverages the inherent semantic consensus between the aligned video and subtitle to achieve weakly-supervised learning. We contribute a new testing set with 3,000 manually-annotated temporal boundaries so that future research on the TEL problem can be quantitatively evaluated. Extensive experiments show the effectiveness of our approach on temporal emotion localization. The repository of this work is at https://github.com/YYJMJC/Temporal-Emotion-Localization-in-Videos.
Emotion Recognition in Conversation (ERC) plays a significant part in Human-Computer Interaction (HCI) systems since it can provide empathetic services. Multimodal ERC can mitigate the drawbacks of uni-modal approaches. Recently, Graph Neural Networks (GNNs) have been widely used in a variety of fields due to their superior performance in relation modeling. In multimodal ERC, GNNs are capable of extracting both long-distance contextual information and inter-modal interactive information. Unfortunately, since existing methods such as MMGCN directly fuse multiple modalities, redundant information may be generated and diverse information may be lost. In this work, we present a directed Graph based Cross-modal Feature Complementation (GraphCFC) module that can efficiently model contextual and interactive information. GraphCFC alleviates the problem of heterogeneity gap in multimodal fusion by utilizing multiple subspace extractors and Pair-wise Cross-modal Complementary (PairCC) strategy. We extract various types of edges from the constructed graph for encoding, thus enabling GNNs to extract crucial contextual and interactive information more accurately when performing message passing. Furthermore, we design a GNN structure called GAT-MLP, which can provide a new unified network framework for multimodal learning. The experimental results on two benchmark datasets show that our GraphCFC outperforms the state-of-the-art (SOTA) approaches.
With the recent proliferation of open textual data on social media platforms, Emotion Detection (ED) from Text has received more attention over the past years. It has many applications, especially for businesses and online service providers, where emotion detection techniques can help them make informed commercial decisions by analyzing customers/users' feelings towards their products and services. In this study, we introduce ArmanEmo, a human-labeled emotion dataset of more than 7000 Persian sentences labeled for seven categories. The dataset has been collected from different resources, including Twitter, Instagram, and Digikala (an Iranian e-commerce company) comments. Labels are based on Ekman's six basic emotions (Anger, Fear, Happiness, Hatred, Sadness, Wonder) and another category (Other) to consider any other emotion not included in Ekman's model. Along with the dataset, we have provided several baseline models for emotion classification focusing on the state-of-the-art transformer-based language models. Our best model achieves a macro-averaged F1 score of 75.39 percent across our test dataset. Moreover, we also conduct transfer learning experiments to compare our proposed dataset's generalization against other Persian emotion datasets. Results of these experiments suggest that our dataset has superior generalizability among the existing Persian emotion datasets. ArmanEmo is publicly available for non-commercial use at https://github.com/Arman-Rayan-Sharif/arman-text-emotion.
With the improved performance of deep learning, the number of studies trying to apply deep learning to human emotion analysis is increasing rapidly. But even with this trend going on, it is still difficult to obtain high-quality images and annotations. For this reason, the Learning from Synthetic Data (LSD) Challenge, which learns from synthetic images and infers from real images, is one of the most interesting areas. In general, Domain Adaptation methods are widely used to address LSD challenges, but there is a limitation that target domains (real images) are still needed. Focusing on these limitations, we propose a framework Bootstrap Your Emotion Latent (BYEL), which uses only synthetic images in training. BYEL is implemented by adding Emotion Classifiers and Emotion Vector Subtraction to the BYOL framework that performs well in Self-Supervised Representation Learning. We train our framework using synthetic images generated from the Aff-wild2 dataset and evaluate it using real images from the Aff-wild2 dataset. The result shows that our framework (0.3084) performs 2.8% higher than the baseline (0.3) on the macro F1 score metric.
Emotion classification of speech and assessment of the emotion strength are required in applications such as emotional text-to-speech and voice conversion. The emotion attribute ranking function based on Support Vector Machine (SVM) was proposed to predict emotion strength for emotional speech corpus. However, the trained ranking function doesn't generalize to new domains, which limits the scope of applications, especially for out-of-domain or unseen speech. In this paper, we propose a data-driven deep learning model, i.e. StrengthNet, to improve the generalization of emotion strength assessment for seen and unseen speech. This is achieved by the fusion of emotional data from various domains. We follow a multi-task learning network architecture that includes an acoustic encoder, a strength predictor, and an auxiliary emotion predictor. Experiments show that the predicted emotion strength of the proposed StrengthNet is highly correlated with ground truth scores for both seen and unseen speech. We release the source codes at: https://github.com/ttslr/StrengthNet.
Document-level relation extraction (RE) aims at extracting relations among entities expressed across multiple sentences, which can be viewed as a multi-label classification problem. In a typical document, most entity pairs do not express any pre-defined relation and are labeled as "none" or "no relation". For good document-level RE performance, it is crucial to distinguish such none class instances (entity pairs) from those of pre-defined classes (relations). However, most existing methods only estimate the probability of pre-defined relations independently without considering the probability of "no relation". This ignores the context of entity pairs and the label correlations between the none class and pre-defined classes, leading to sub-optimal predictions. To address this problem, we propose a new multi-label loss that encourages large margins of label confidence scores between each pre-defined class and the none class, which enables captured label correlations and context-dependent thresholding for label prediction. To gain further robustness against positive-negative imbalance and mislabeled data that could appear in real-world RE datasets, we propose a margin regularization and a margin shifting technique. Experimental results demonstrate that our method significantly outperforms existing multi-label losses for document-level RE and works well in other multi-label tasks such as emotion classification when none class instances are available for training.
Electroencephalography (EEG) is shown to be a valuable data source for evaluating subjects' mental states. However, the interpretation of multi-modal EEG signals is challenging, as they suffer from poor signal-to-noise-ratio, are highly subject-dependent, and are bound to the equipment and experimental setup used, (i.e. domain). This leads to machine learning models often suffer from poor generalization ability, where they perform significantly worse on real-world data than on the exploited training data. Recent research heavily focuses on cross-subject and cross-session transfer learning frameworks to reduce domain calibration efforts for EEG signals. We argue that multi-source learning via learning domain-invariant representations from multiple data-sources is a viable alternative, as the available data from different EEG data-source domains (e.g., subjects, sessions, experimental setups) grow massively. We propose an adversarial inference approach to learn data-source invariant representations in this context, enabling multi-source learning for EEG-based brain-computer interfaces. We unify EEG recordings from different source domains (i.e., emotion recognition datasets SEED, SEED-IV, DEAP, DREAMER), and demonstrate the feasibility of our invariant representation learning approach in suppressing data-source-relevant information leakage by 35% while still achieving stable EEG-based emotion classification performance.
Humans express their emotions via facial expressions, voice intonation and word choices. To infer the nature of the underlying emotion, recognition models may use a single modality, such as vision, audio, and text, or a combination of modalities. Generally, models that fuse complementary information from multiple modalities outperform their uni-modal counterparts. However, a successful model that fuses modalities requires components that can effectively aggregate task-relevant information from each modality. As cross-modal attention is seen as an effective mechanism for multi-modal fusion, in this paper we quantify the gain that such a mechanism brings compared to the corresponding self-attention mechanism. To this end, we implement and compare a cross-attention and a self-attention model. In addition to attention, each model uses convolutional layers for local feature extraction and recurrent layers for global sequential modelling. We compare the models using different modality combinations for a 7-class emotion classification task using the IEMOCAP dataset. Experimental results indicate that albeit both models improve upon the state-of-the-art in terms of weighted and unweighted accuracy for tri- and bi-modal configurations, their performance is generally statistically comparable. The code to replicate the experiments is available at https://github.com/smartcameras/SelfCrossAttn
We propose PARSE, a novel semi-supervised architecture for learning strong EEG representations for emotion recognition. To reduce the potential distribution mismatch between the large amounts of unlabeled data and the limited amount of labeled data, PARSE uses pairwise representation alignment. First, our model performs data augmentation followed by label guessing for large amounts of original and augmented unlabeled data. This is then followed by sharpening of the guessed labels and convex combinations of the unlabeled and labeled data. Finally, representation alignment and emotion classification are performed. To rigorously test our model, we compare PARSE to several state-of-the-art semi-supervised approaches which we implement and adapt for EEG learning. We perform these experiments on four public EEG-based emotion recognition datasets, SEED, SEED-IV, SEED-V and AMIGOS (valence and arousal). The experiments show that our proposed framework achieves the overall best results with varying amounts of limited labeled samples in SEED, SEED-IV and AMIGOS (valence), while approaching the overall best result (reaching the second-best) in SEED-V and AMIGOS (arousal). The analysis shows that our pairwise representation alignment considerably improves the performance by reducing the distribution alignment between unlabeled and labeled data, especially when only 1 sample per class is labeled.
Deep learning based electroencephalography (EEG) signal processing methods are known to suffer from poor test-time generalization due to the changes in data distribution. This becomes a more challenging problem when privacy-preserving representation learning is of interest such as in clinical settings. To that end, we propose a multi-source learning architecture where we extract domain-invariant representations from dataset-specific private encoders. Our model utilizes a maximum-mean-discrepancy (MMD) based domain alignment approach to impose domain-invariance for encoded representations, which outperforms state-of-the-art approaches in EEG-based emotion classification. Furthermore, representations learned in our pipeline preserve domain privacy as dataset-specific private encoding alleviates the need for conventional, centralized EEG-based deep neural network training approaches with shared parameters.
The portrayal of negative emotions such as anger can vary widely between cultures and contexts, depending on the acceptability of expressing full-blown emotions rather than suppression to maintain harmony. The majority of emotional datasets collect data under the broad label ``anger", but social signals can range from annoyed, contemptuous, angry, furious, hateful, and more. In this work, we curated the first in-the-wild multicultural video dataset of emotions, and deeply explored anger-related emotional expressions by asking culture-fluent annotators to label the videos with 6 labels and 13 emojis in a multi-label framework. We provide a baseline multi-label classifier on our dataset, and show how emojis can be effectively used as a language-agnostic tool for annotation.
Several recent studies on dyadic human-human interactions have been done on conversations without specific business objectives. However, many companies might benefit from studies dedicated to more precise environments such as after sales services or customer satisfaction surveys. In this work, we place ourselves in the scope of a live chat customer service in which we want to detect emotions and their evolution in the conversation flow. This context leads to multiple challenges that range from exploiting restricted, small and mostly unlabeled datasets to finding and adapting methods for such context.We tackle these challenges by using Few-Shot Learning while making the hypothesis it can serve conversational emotion classification for different languages and sparse labels. We contribute by proposing a variation of Prototypical Networks for sequence labeling in conversation that we name ProtoSeq. We test this method on two datasets with different languages: daily conversations in English and customer service chat conversations in French. When applied to emotion classification in conversations, our method proved to be competitive even when compared to other ones.
Fine-grained classification involves dealing with datasets with larger number of classes with subtle differences between them. Guiding the model to focus on differentiating dimensions between these commonly confusable classes is key to improving performance on fine-grained tasks. In this work, we analyse the contrastive fine-tuning of pre-trained language models on two fine-grained text classification tasks, emotion classification and sentiment analysis. We adaptively embed class relationships into a contrastive objective function to help differently weigh the positives and negatives, and in particular, weighting closely confusable negatives more than less similar negative examples. We find that Label-aware Contrastive Loss outperforms previous contrastive methods, in the presence of larger number and/or more confusable classes, and helps models to produce output distributions that are more differentiated.
Identifying emotions from text is crucial for a variety of real world tasks. We consider the two largest now-available corpora for emotion classification: GoEmotions, with 58k messages labelled by readers, and Vent, with 33M writer-labelled messages. We design a benchmark and evaluate several feature spaces and learning algorithms, including two simple yet novel models on top of BERT that outperform previous strong baselines on GoEmotions. Through an experiment with human participants, we also analyze the differences between how writers express emotions and how readers perceive them. Our results suggest that emotions expressed by writers are harder to identify than emotions that readers perceive. We share a public web interface for researchers to explore our models.
Existing audio-language task-specific predictive approaches focus on building complicated late-fusion mechanisms. However, these models are facing challenges of overfitting with limited labels and low model generalization abilities. In this paper, we present a Cross-modal Transformer for Audio-and-Language, i.e., CTAL, which aims to learn the intra-modality and inter-modality connections between audio and language through two proxy tasks on a large amount of audio-and-language pairs: masked language modeling and masked cross-modal acoustic modeling. After fine-tuning our pre-trained model on multiple downstream audio-and-language tasks, we observe significant improvements across various tasks, such as, emotion classification, sentiment analysis, and speaker verification. On this basis, we further propose a specially-designed fusion mechanism that can be used in fine-tuning phase, which allows our pre-trained model to achieve better performance. Lastly, we demonstrate detailed ablation studies to prove that both our novel cross-modality fusion component and audio-language pre-training methods significantly contribute to the promising results.
This article presents a benchmark study of symbolic piano music classification using the masked language modelling approach of the Bidirectional Encoder Representations from Transformers (BERT). Specifically, we consider two types of MIDI data: MIDI scores, which are musical scores rendered directly into MIDI with no dynamics and precisely aligned with the metrical grid notated by its composer and MIDI performances, which are MIDI encodings of human performances of musical scoresheets. With five public-domain datasets of single-track piano MIDI files, we pre-train two 12-layer Transformer models using the BERT approach, one for MIDI scores and the other for MIDI performances, and fine-tune them for four downstream classification tasks. These include two note-level classification tasks (melody extraction and velocity prediction) and two sequence-level classification tasks (style classification and emotion classification). Our evaluation shows that the BERT approach leads to higher classification accuracy than recurrent neural network (RNN)-based baselines.
Symbolic music understanding, which refers to the understanding of music from the symbolic data (e.g., MIDI format, but not audio), covers many music applications such as genre classification, emotion classification, and music pieces matching. While good music representations are beneficial for these applications, the lack of training data hinders representation learning. Inspired by the success of pre-training models in natural language processing, in this paper, we develop MusicBERT, a large-scale pre-trained model for music understanding. To this end, we construct a large-scale symbolic music corpus that contains more than 1 million music songs. Since symbolic music contains more structural (e.g., bar, position) and diverse information (e.g., tempo, instrument, and pitch), simply adopting the pre-training techniques from NLP to symbolic music only brings marginal gains. Therefore, we design several mechanisms, including OctupleMIDI encoding and bar-level masking strategy, to enhance pre-training with symbolic music data. Experiments demonstrate the advantages of MusicBERT on four music understanding tasks, including melody completion, accompaniment suggestion, genre classification, and style classification. Ablation studies also verify the effectiveness of our designs of OctupleMIDI encoding and bar-level masking strategy in MusicBERT.
Emotion recognition from speech is a challenging task. Re-cent advances in deep learning have led bi-directional recur-rent neural network (Bi-RNN) and attention mechanism as astandard method for speech emotion recognition, extractingand attending multi-modal features - audio and text, and thenfusing them for downstream emotion classification tasks. Inthis paper, we propose a simple yet efficient neural networkarchitecture to exploit both acoustic and lexical informationfrom speech. The proposed framework using multi-scale con-volutional layers (MSCNN) to obtain both audio and text hid-den representations. Then, a statistical pooling unit (SPU)is used to further extract the features in each modality. Be-sides, an attention module can be built on top of the MSCNN-SPU (audio) and MSCNN (text) to further improve the perfor-mance. Extensive experiments show that the proposed modeloutperforms previous state-of-the-art methods on IEMOCAPdataset with four emotion categories (i.e., angry, happy, sadand neutral) in both weighted accuracy (WA) and unweightedaccuracy (UA), with an improvement of 5.0% and 5.2% respectively under the ASR setting.