Studying under-represented music traditions under the MIR scope is crucial, not only for developing novel analysis tools, but also for unveiling musical functions that might prove useful in studying world musics. This paper presents a dataset for Greek Traditional and Folk music that includes 1570 pieces, summing in around 80 hours of data. The dataset incorporates YouTube timestamped links for retrieving audio and video, along with rich metadata information with regards to instrumentation, geography and genre, among others. The content has been collected from a Greek documentary series that is available online, where academics present music traditions of Greece with live music and dance performance during the show, along with discussions about social, cultural and musicological aspects of the presented music. Therefore, this procedure has resulted in a significant wealth of descriptions regarding a variety of aspects, such as musical genre, places of origin and musical instruments. In addition, the audio recordings were performed under strict production-level specifications, in terms of recording equipment, leading to very clean and homogeneous audio content. In this work, apart from presenting the dataset in detail, we propose a baseline deep-learning classification approach to recognize the involved musicological attributes. The dataset, the baseline classification methods and the models are provided in public repositories. Future directions for further refining the dataset are also discussed.
We propose a novel deep architecture for the task of reasoning about social interactions in videos. We leverage the multi-step reasoning capabilities of Compositional Attention Networks (MAC), and propose a multimodal extension (MAC-X). MAC-X is based on a recurrent cell that performs iterative mid-level fusion of input modalities (visual, auditory, text) over multiple reasoning steps, by use of a temporal attention mechanism. We then combine MAC-X with LSTMs for temporal input processing in an end-to-end architecture. Our ablation studies show that the proposed MAC-X architecture can effectively leverage multimodal input cues using mid-level fusion mechanisms. We apply MAC-X to the task of Social Video Question Answering in the Social IQ dataset and obtain a 2.5% absolute improvement in terms of binary accuracy over the current state-of-the-art.
Task-oriented dialogue systems often employ a Dialogue State Tracker (DST) to successfully complete conversations. Recent state-of-the-art DST implementations rely on schemata of diverse services to improve model robustness and handle zero-shot generalization to new domains [1], however such methods [2, 3] typically require multiple large scale transformer models and long input sequences to perform well. We propose a single multi-task BERT-based model that jointly solves the three DST tasks of intent prediction, requested slot prediction and slot filling. Moreover, we propose an efficient and parsimonious encoding of the dialogue history and service schemata that is shown to further improve performance. Evaluation on the SGD dataset shows that our approach outperforms the baseline SGP-DST by a large margin and performs well compared to the state-of-the-art, while being significantly more computationally efficient. Extensive ablation studies are performed to examine the contributing factors to the success of our model.
Recent deep learning Text-to-Speech (TTS) systems have achieved impressive performance by generating speech close to human parity. However, they suffer from training stability issues as well as incorrect alignment of the intermediate acoustic representation with the input text sequence. In this work, we introduce Regotron, a regularized version of Tacotron2 which aims to alleviate the training issues and at the same time produce monotonic alignments. Our method augments the vanilla Tacotron2 objective function with an additional term, which penalizes non-monotonic alignments in the location-sensitive attention mechanism. By properly adjusting this regularization term we show that the loss curves become smoother, and at the same time Regotron consistently produces monotonic alignments in unseen examples even at an early stage (13\% of the total number of epochs) of its training process, whereas the fully converged Tacotron2 fails to do so. Moreover, our proposed regularization method has no additional computational overhead, while reducing common TTS mistakes and achieving slighlty improved speech naturalness according to subjective mean opinion scores (MOS) collected from 50 evaluators.
Current deep learning approaches for multimodal fusion rely on bottom-up fusion of high and mid-level latent modality representations (late/mid fusion) or low level sensory inputs (early fusion). Models of human perception highlight the importance of top-down fusion, where high-level representations affect the way sensory inputs are perceived, i.e. cognition affects perception. These top-down interactions are not captured in current deep learning models. In this work we propose a neural architecture that captures top-down cross-modal interactions, using a feedback mechanism in the forward pass during network training. The proposed mechanism extracts high-level representations for each modality and uses these representations to mask the sensory inputs, allowing the model to perform top-down feature masking. We apply the proposed model for multimodal sentiment recognition on CMU-MOSEI. Our method shows consistent improvements over the well established MulT and over our strong late fusion baseline, achieving state-of-the-art results.
In this paper, we introduce EmpBot: an end-to-end empathetic chatbot. Empathetic conversational agents should not only understand what is being discussed, but also acknowledge the implied feelings of the conversation partner and respond appropriately. To this end, we propose a method based on a transformer pretrained language model (T5). Specifically, during finetuning we propose to use three objectives: response language modeling, sentiment understanding, and empathy forcing. The first objective is crucial for generating relevant and coherent responses, while the next ones are significant for acknowledging the sentimental state of the conversational partner and for favoring empathetic responses. We evaluate our model on the EmpatheticDialogues dataset using both automated metrics and human evaluation. The inclusion of the sentiment understanding and empathy forcing auxiliary losses favor empathetic responses, as human evaluation results indicate, comparing with the current state-of-the-art.
In this work we explore Unsupervised Domain Adaptation (UDA) of pretrained language models for downstream tasks. We introduce UDALM, a fine-tuning procedure, using a mixed classification and Masked Language Model loss, that can adapt to the target domain distribution in a robust and sample efficient manner. Our experiments show that performance of models trained with the mixed loss scales with the amount of available target data and the mixed loss can be effectively used as a stopping criterion during UDA training. Furthermore, we discuss the relationship between A-distance and the target error and explore some limitations of the Domain Adversarial Training approach. Our method is evaluated on twelve domain pairs of the Amazon Reviews Sentiment dataset, yielding $91.74\%$ accuracy, which is an $1.11\%$ absolute improvement over the state-of-the-art.
We examine the use of linear and non-linear dimensionality reduction algorithms for extracting low-rank feature representations for speech emotion recognition. Two feature sets are used, one based on low-level descriptors and their aggregations (IS10) and one modeling recurrence dynamics of speech (RQA), as well as their fusion. We report speech emotion recognition (SER) results for learned representations on two databases using different classification methods. Classification with low-dimensional representations yields performance improvement in a variety of settings. This indicates that dimensionality reduction is an effective way to combat the curse of dimensionality for SER. Visualization of features in two dimensions provides insight into discriminatory abilities of reduced feature sets.
Contemporary state-of-the-art approaches to Zero-Shot Learning (ZSL) train generative nets to synthesize examples conditioned on the provided metadata. Thereafter, classifiers are trained on these synthetic data in a supervised manner. In this work, we introduce Z2FSL, an end-to-end generative ZSL framework that uses such an approach as a backbone and feeds its synthesized output to a Few-Shot Learning (FSL) algorithm. The two modules are trained jointly. Z2FSL solves the ZSL problem with a FSL algorithm, reducing, in effect, ZSL to FSL. A wide class of algorithms can be integrated within our framework. Our experimental results show consistent improvement over several baselines. The proposed method, evaluated across standard benchmarks, shows state-of-the-art or competitive performance in ZSL and Generalized ZSL tasks.
In this paper, we present a novel approach for incorporating external knowledge in Recurrent Neural Networks (RNNs). We propose the integration of lexicon features into the self-attention mechanism of RNN-based architectures. This form of conditioning on the attention distribution, enforces the contribution of the most salient words for the task at hand. We introduce three methods, namely attentional concatenation, feature-based gating and affine transformation. Experiments on six benchmark datasets show the effectiveness of our methods. Attentional feature-based gating yields consistent performance improvement across tasks. Our approach is implemented as a simple add-on module for RNN-based models with minimal computational overhead and can be adapted to any deep neural architecture.