Multimodal sentiment analysis has attracted increasing attention and lots of models have been proposed. However, the performance of the state-of-the-art models decreases sharply when they are deployed in the real world. We find that the main reason is that real-world applications can only access the text outputs by the automatic speech recognition (ASR) models, which may be with errors because of the limitation of model capacity. Through further analysis of the ASR outputs, we find that in some cases the sentiment words, the key sentiment elements in the textual modality, are recognized as other words, which makes the sentiment of the text change and hurts the performance of multimodal sentiment models directly. To address this problem, we propose the sentiment word aware multimodal refinement model (SWRM), which can dynamically refine the erroneous sentiment words by leveraging multimodal sentiment clues. Specifically, we first use the sentiment word position detection module to obtain the most possible position of the sentiment word in the text and then utilize the multimodal sentiment word refinement module to dynamically refine the sentiment word embeddings. The refined embeddings are taken as the textual inputs of the multimodal feature fusion module to predict the sentiment labels. We conduct extensive experiments on the real-world datasets including MOSI-Speechbrain, MOSI-IBM, and MOSI-iFlytek and the results demonstrate the effectiveness of our model, which surpasses the current state-of-the-art models on three datasets. Furthermore, our approach can be adapted for other multimodal feature fusion models easily. Data and code are available at https://github.com/albertwy/SWRM.
Recurrent neural networks such as Long Short-Term Memories (LSTMs) learn temporal dependencies by keeping an internal state, making them ideal for time-series problems such as speech recognition. However, the output-to-input feedback creates distinctive memory bandwidth and scalability challenges in designing accelerators for RNNs. We present Muntaniala, an RNN accelerator architecture for LSTM inference with a silicon-measured energy-efficiency of 3.25$TOP/s/W$ and performance of 30.53$GOP/s$ in UMC 65 $nm$ technology. The scalable design of Muntaniala allows running large RNN models by combining multiple tiles in a systolic array. We keep all parameters stationary on every die in the array, drastically reducing the I/O communication to only loading new features and sharing partial results with other dies. For quantifying the overall system power, including I/O power, we built Vau da Muntanialas, to the best of our knowledge, the first demonstration of a systolic multi-chip-on-PCB array of RNN accelerator. Our multi-die prototype performs LSTM inference with 192 hidden states in 330$\mu s$ with a total system power of 9.0$mW$ at 10$MHz$ consuming 2.95$\mu J$. Targeting the 8/16-bit quantization implemented in Muntaniala, we show a phoneme error rate (PER) drop of approximately 3% with respect to floating-point (FP) on a 3L-384NH-123NI LSTM network on the TIMIT dataset.
Learning disentangled representations of high-dimensional data is currently an active research area. However, compared to the field of computer vision, less work has been done for speech processing. In this paper, we provide a review of two representative efforts on this topic and propose the novel concept of fine-grained disentangled speech representation learning.
The speech chain mechanism integrates automatic speech recognition (ASR) and text-to-speech synthesis (TTS) modules into a single cycle during training. In our previous work, we applied a speech chain mechanism as a semi-supervised learning. It provides the ability for ASR and TTS to assist each other when they receive unpaired data and let them infer the missing pair and optimize the model with reconstruction loss. If we only have speech without transcription, ASR generates the most likely transcription from the speech data, and then TTS uses the generated transcription to reconstruct the original speech features. However, in previous papers, we just limited our back-propagation to the closest module, which is the TTS part. One reason is that back-propagating the error through the ASR is challenging due to the output of the ASR are discrete tokens, creating non-differentiability between the TTS and ASR. In this paper, we address this problem and describe how to thoroughly train a speech chain end-to-end for reconstruction loss using a straight-through estimator (ST). Experimental results revealed that, with sampling from ST-Gumbel-Softmax, we were able to update ASR parameters and improve the ASR performances by 11\% relative CER reduction compared to the baseline.
In this paper, we present a novel two-pass approach to unify streaming and non-streaming end-to-end (E2E) speech recognition in a single model. Our model adopts the hybrid CTC/attention architecture, in which the conformer layers in the encoder are modified. We propose a dynamic chunk-based attention strategy to allow arbitrary right context length. At inference time, the CTC decoder generates n-best hypotheses in a streaming way. The inference latency could be easily controlled by only changing the chunk size. The CTC hypotheses are then rescored by the attention decoder to get the final result. This efficient rescoring process causes very little sentence-level latency. Our experiments on the open 170-hour AISHELL-1 dataset show that, the proposed method can unify the streaming and non-streaming model simply and efficiently. On the AISHELL-1 test set, our unified model achieves 5.60% relative character error rate (CER) reduction in non-streaming ASR compared to a standard non-streaming transformer. The same model achieves 5.42% CER with 640ms latency in a streaming ASR system.
We investigate the effect and usefulness of spontaneity (i.e. whether a given speech is spontaneous or not) in speech in the context of emotion recognition. We hypothesize that emotional content in speech is interrelated with its spontaneity, and use spontaneity classification as an auxiliary task to the problem of emotion recognition. We propose two supervised learning settings that utilize spontaneity to improve speech emotion recognition: a hierarchical model that performs spontaneity detection before performing emotion recognition, and a multitask learning model that jointly learns to recognize both spontaneity and emotion. Through various experiments on the well known IEMOCAP database, we show that by using spontaneity detection as an additional task, significant improvement can be achieved over emotion recognition systems that are unaware of spontaneity. We achieve state-of-the-art emotion recognition accuracy (4-class, 69.1%) on the IEMOCAP database outperforming several relevant and competitive baselines.
Recent research has shown that word embedding spaces learned from text corpora of different languages can be aligned without any parallel data supervision. Inspired by the success in unsupervised cross-lingual word embeddings, in this paper we target learning a cross-modal alignment between the embedding spaces of speech and text learned from corpora of their respective modalities in an unsupervised fashion. The proposed framework learns the individual speech and text embedding spaces, and attempts to align the two spaces via adversarial training, followed by a refinement procedure. We show how our framework could be used to perform spoken word classification and translation, and the results on these two tasks demonstrate that the performance of our unsupervised alignment approach is comparable to its supervised counterpart. Our framework is especially useful for developing automatic speech recognition (ASR) and speech-to-text translation systems for low- or zero-resource languages, which have little parallel audio-text data for training modern supervised ASR and speech-to-text translation models, but account for the majority of the languages spoken across the world.
In this paper, we propose to use pre-trained features from end-to-end ASR models to solve the speech sentiment analysis problem as a down-stream task. We show that end-to-end ASR features, which integrate both acoustic and text information from speech, achieve promising results. We use RNN with self-attention as the sentiment classifier, which also provides an easy visualization through attention weights to help interpret model predictions. We use well benchmarked IEMOCAP dataset and a new large-scale sentiment analysis dataset SWBD-senti for evaluation. Our approach improves the-state-of-the-art accuracy on IEMOCAP from 66.6% to 71.7%, and achieves an accuracy of 70.10% on SWBD-senti with more than 49,500 utterances.
This paper presents sampling-based speech parameter generation using moment-matching networks for Deep Neural Network (DNN)-based speech synthesis. Although people never produce exactly the same speech even if we try to express the same linguistic and para-linguistic information, typical statistical speech synthesis produces completely the same speech, i.e., there is no inter-utterance variation in synthetic speech. To give synthetic speech natural inter-utterance variation, this paper builds DNN acoustic models that make it possible to randomly sample speech parameters. The DNNs are trained so that they make the moments of generated speech parameters close to those of natural speech parameters. Since the variation of speech parameters is compressed into a low-dimensional simple prior noise vector, our algorithm has lower computation cost than direct sampling of speech parameters. As the first step towards generating synthetic speech that has natural inter-utterance variation, this paper investigates whether or not the proposed sampling-based generation deteriorates synthetic speech quality. In evaluation, we compare speech quality of conventional maximum likelihood-based generation and proposed sampling-based generation. The result demonstrates the proposed generation causes no degradation in speech quality.
Continuous affect prediction involves the discrete time-continuous regression of affect dimensions. Dimensions to be predicted often include arousal and valence. Continuous affect prediction researchers are now embracing multimodal model input. This provides motivation for researchers to investigate previously unexplored affective cues. Speech-based cues have traditionally received the most attention for affect prediction, however, non-verbal inputs have significant potential to increase the performance of affective computing systems and in addition, allow affect modelling in the absence of speech. However, non-verbal inputs that have received little attention for continuous affect prediction include eye and head-based cues. The eyes are involved in emotion displays and perception while head-based cues have been shown to contribute to emotion conveyance and perception. Additionally, these cues can be estimated non-invasively from video, using modern computer vision tools. This work exploits this gap by comprehensively investigating head and eye-based features and their combination with speech for continuous affect prediction. Hand-crafted, automatically generated and CNN-learned features from these modalities will be investigated for continuous affect prediction. The highest performing feature sets and feature set combinations will answer how effective these features are for the prediction of an individual's affective state.