Abstract:Non-autoregressive automatic speech recognition (ASR) modeling has received increasing attention recently because of its fast decoding speed and superior performance. Among representatives, methods based on the connectionist temporal classification (CTC) are still a dominating stream. However, the theoretically inherent flaw, the assumption of independence between tokens, creates a performance barrier for the school of works. To mitigate the challenge, we propose a context-aware knowledge transferring strategy, consisting of a knowledge transferring module and a context-aware training strategy, for CTC-based ASR. The former is designed to distill linguistic information from a pre-trained language model, and the latter is framed to modulate the limitations caused by the conditional independence assumption. As a result, a knowledge-injected context-aware CTC-based ASR built upon the wav2vec2.0 is presented in this paper. A series of experiments on the AISHELL-1 and AISHELL-2 datasets demonstrate the effectiveness of the proposed method.
Abstract:In this paper, inspired by the successes of visionlanguage pre-trained models and the benefits from training with adversarial attacks, we present a novel transformerbased cross-modal fusion modeling by incorporating the both notions for VQA challenge 2021. Specifically, the proposed model is on top of the architecture of VinVL model [19], and the adversarial training strategy [4] is applied to make the model robust and generalized. Moreover, two implementation tricks are also used in our system to obtain better results. The experiments demonstrate that the novel framework can achieve 76.72% on VQAv2 test-std set.
Abstract:Transformer-based models have led to a significant innovation in various classic and practical subjects, including speech processing, natural language processing, and computer vision. On top of the transformer, the attention-based end-to-end automatic speech recognition (ASR) models have become a popular fashion in recent years. Specifically, the non-autoregressive modeling, which can achieve fast inference speed and comparable performance when compared to conventional autoregressive methods, is an emergent research topic. In the context of natural language processing, the bidirectional encoder representations from transformers (BERT) model has received widespread attention, partially due to its ability to infer contextualized word representations and to obtain superior performances of downstream tasks by performing only simple fine-tuning. In order to not only inherit the advantages of non-autoregressive ASR modeling, but also receive benefits from a pre-trained language model (e.g., BERT), a non-autoregressive transformer-based end-to-end ASR model based on BERT is presented in this paper. A series of experiments conducted on the AISHELL-1 dataset demonstrates competitive or superior results of the proposed model when compared to state-of-the-art ASR systems.
Abstract:We propose a simple method for automatic speech recognition (ASR) by fine-tuning BERT, which is a language model (LM) trained on large-scale unlabeled text data and can generate rich contextual representations. Our assumption is that given a history context sequence, a powerful LM can narrow the range of possible choices and the speech signal can be used as a simple clue. Hence, comparing to conventional ASR systems that train a powerful acoustic model (AM) from scratch, we believe that speech recognition is possible by simply fine-tuning a BERT model. As an initial study, we demonstrate the effectiveness of the proposed idea on the AISHELL dataset and show that stacking a very simple AM on top of BERT can yield reasonable performance.
Abstract:Conventional seq2seq chatbot models attempt only to find sentences with the highest probabilities conditioned on the input sequences, without considering the sentiment of the output sentences. In this paper, we investigate four models to scale or adjust the sentiment of the chatbot response: a persona-based model, reinforcement learning, a plug and play model, and CycleGAN, all based on the seq2seq model. We also develop machine-evaluated metrics to estimate whether the responses are reasonable given the input. These metrics, together with human evaluation, are used to analyze the performance of the four models in terms of different aspects; reinforcement learning and CycleGAN are shown to be very attractive.
Abstract:In a spoken multiple-choice question answering (SMCQA) task, given a passage, a question, and multiple choices all in the form of speech, the machine needs to pick the correct choice to answer the question. While the audio could contain useful cues for SMCQA, usually only the auto-transcribed text is utilized in system development. Thanks to the large-scaled pre-trained language representation models, such as the bidirectional encoder representations from transformers (BERT), systems with only auto-transcribed text can still achieve a certain level of performance. However, previous studies have evidenced that acoustic-level statistics can offset text inaccuracies caused by the automatic speech recognition systems or representation inadequacy lurking in word embedding generators, thereby making the SMCQA system robust. Along the line of research, this study concentrates on designing a BERT-based SMCQA framework, which not only inherits the advantages of contextualized language representations learned by BERT, but integrates the complementary acoustic-level information distilled from audio with the text-level information. Consequently, an audio-enriched BERT-based SMCQA framework is proposed. A series of experiments demonstrates remarkable improvements in accuracy over selected baselines and SOTA systems on a published Chinese SMCQA dataset.
Abstract:Recent developments in deep learning have led to a significant innovation in various classic and practical subjects, including speech recognition, computer vision, question answering, information retrieval and so on. In the context of natural language processing (NLP), language representations have shown giant successes in many downstream tasks, so the school of studies have become a major stream of research recently. Because the immenseness of multimedia data along with speech have spread around the world in our daily life, spoken document retrieval (SDR) has become an important research subject in the past decades. Targeting on enhancing the SDR performance, the paper concentrates on proposing a neural retrieval framework, which assembles the merits of using language modeling (LM) mechanism in SDR and leveraging the abstractive information learned by the language representation models. Consequently, to our knowledge, this is a pioneer study on supervised training of a neural LM-based SDR framework, especially combined with the pretrained language representation methods.
Abstract:Producing a large annotated speech corpus for training ASR systems remains difficult for more than 95% of languages all over the world which are low-resourced, but collecting a relatively big unlabeled data set for such languages is more achievable. This is why some initial effort have been reported on completely unsupervised speech recognition learned from unlabeled data only, although with relatively high error rates. In this paper, we develop a Generative Adversarial Network (GAN) to achieve this purpose, in which a Generator and a Discriminator learn from each other iteratively to improve the performance. We further use a set of Hidden Markov Models (HMMs) iteratively refined from the machine generated labels to work in harmony with the GAN. The initial experiments on TIMIT data set achieve an phone error rate of 33.1%, which is 8.5% lower than the previous state-of-the-art.
Abstract:Conventional seq2seq chatbot models only try to find the sentences with the highest probabilities conditioned on the input sequences, without considering the sentiment of the output sentences. Some research works trying to modify the sentiment of the output sequences were reported. In this paper, we propose five models to scale or adjust the sentiment of the chatbot response: persona-based model, reinforcement learning, plug and play model, sentiment transformation network and cycleGAN, all based on the conventional seq2seq model. We also develop two evaluation metrics to estimate if the responses are reasonable given the input. These metrics together with other two popularly used metrics were used to analyze the performance of the five proposed models on different aspects, and reinforcement learning and cycleGAN were shown to be very attractive. The evaluation metrics were also found to be well correlated with human evaluation.
Abstract:Unsupervised discovery of acoustic tokens from audio corpora without annotation and learning vector representations for these tokens have been widely studied. Although these techniques have been shown successful in some applications such as query-by-example Spoken Term Detection (STD), the lack of mapping relationships between these discovered tokens and real phonemes have limited the down-stream applications. This paper represents probably the first attempt towards the goal of completely unsupervised phoneme recognition, or mapping audio signals to phoneme sequences without phoneme-labeled audio data. The basic idea is to cluster the embedded acoustic tokens and learn the mapping between the cluster sequences and the unknown phoneme sequences with a Generative Adversarial Network (GAN). An unsupervised phoneme recognition accuracy of 36% was achieved in the preliminary experiments.