Improving Mandarin End-to-End Speech Recognition with Word N-gram Language Model

Despite the rapid progress of end-to-end (E2E) automatic speech recognition (ASR), it has been shown that incorporating external language models (LMs) into the decoding can further improve the recognition performance of E2E ASR systems. To align with the modeling units adopted in E2E ASR systems, subword-level (e.g., characters, BPE) LMs are usually used to cooperate with current E2E ASR systems. However, the use of subword-level LMs will ignore the word-level information, which may limit the strength of the external LMs in E2E ASR. Although several methods have been proposed to incorporate word-level external LMs in E2E ASR, these methods are mainly designed for languages with clear word boundaries such as English and cannot be directly applied to languages like Mandarin, in which each character sequence can have multiple corresponding word sequences. To this end, we propose a novel decoding algorithm where a word-level lattice is constructed on-the-fly to consider all possible word sequences for each partial hypothesis. Then, the LM score of the hypothesis is obtained by intersecting the generated lattice with an external word N-gram LM. The proposed method is examined on both Attention-based Encoder-Decoder (AED) and Neural Transducer (NT) frameworks. Experiments suggest that our method consistently outperforms subword-level LMs, including N-gram LM and neural network LM. We achieve state-of-the-art results on both Aishell-1 (CER 4.18%) and Aishell-2 (CER 5.06%) datasets and reduce CER by 14.8% relatively on a 21K-hour Mandarin dataset.

Consistent Training and Decoding For End-to-end Speech Recognition Using Lattice-free MMI

Recently, End-to-End (E2E) frameworks have achieved remarkable results on various Automatic Speech Recognition (ASR) tasks. However, Lattice-Free Maximum Mutual Information (LF-MMI), as one of the discriminative training criteria that show superior performance in hybrid ASR systems, is rarely adopted in E2E ASR frameworks. In this work, we propose a novel approach to integrate LF-MMI criterion into E2E ASR frameworks in both training and decoding stages. The proposed approach shows its effectiveness on two of the most widely used E2E frameworks including Attention-Based Encoder-Decoders (AEDs) and Neural Transducers (NTs). Experiments suggest that the introduction of the LF-MMI criterion consistently leads to significant performance improvements on various datasets and different E2E ASR frameworks. The best of our models achieves competitive CER of 4.1\% / 4.4\% on Aishell-1 dev/test set; we also achieve significant error reduction on Aishell-2 and Librispeech datasets over strong baselines.

Detect what you want: Target Sound Detection

Human beings can perceive a target sound that we are interested in from a multi-source environment by the selective auditory attention, however, such functionality was hardly ever explored in machine hearing.This paper address the target sound detection (TSD), which aims to detect the target sound signal from a mixture audio when a target sound's reference audio is given.We present a novel target sound detection network (TSDNet) which consists of two main parts: A conditional and a detection network. The former aims at generating a sound-discriminative conditional embedding vector representing the global information of the target sound. The latter takes both the mixture audio and the conditional embedding vector as inputs, and produces the detection result. These two networks can be jointly optimized with a multi-task learning approach to further improve the performance. In addition, we study both supervised and weakly supervised strategies to train TSDNet.To evaluate our methods, we build a target sound detection dataset (TSD Dataset) based on URBAN-SED and URBAN-SOUND8K datasets. Experimental results indicate our system can get better performance than universal sound event detection.

CLIP Meets Video Captioners: Attribute-Aware Representation Learning Promotes Accurate Captioning

For video captioning, "pre-training and fine-tuning" has become a de facto paradigm, where ImageNet Pre-training (INP) is usually used to help encode the video content, and a task-oriented network is fine-tuned from scratch to cope with caption generation. Comparing INP with the recently proposed CLIP (Contrastive Language-Image Pre-training), this paper investigates the potential deficiencies of INP for video captioning and explores the key to generating accurate descriptions. Specifically, our empirical study on INP vs. CLIP shows that INP makes video caption models tricky to capture attributes' semantics and sensitive to irrelevant background information. By contrast, CLIP's significant boost in caption quality highlights the importance of attribute-aware representation learning. We are thus motivated to introduce Dual Attribute Prediction, an auxiliary task requiring a video caption model to learn the correspondence between video content and attributes and the co-occurrence relations between attributes. Extensive experiments on benchmark datasets demonstrate that our approach enables better learning of attribute-aware representations, bringing consistent improvements on models with different architectures and decoding algorithms.

Improving the Performance of Automated Audio Captioning via Integrating the Acoustic and Semantic Information

Automated audio captioning (AAC) has developed rapidly in recent years, involving acoustic signal processing and natural language processing to generate human-readable sentences for audio clips. The current models are generally based on the neural encoder-decoder architecture, and their decoder mainly uses acoustic information that is extracted from the CNN-based encoder. However, they have ignored semantic information that could help the AAC model to generate meaningful descriptions. This paper proposes a novel approach for automated audio captioning based on incorporating semantic and acoustic information. Specifically, our audio captioning model consists of two sub-modules. (1) The pre-trained keyword encoder utilizes pre-trained ResNet38 to initialize its parameters, and then it is trained by extracted keywords as labels. (2) The multi-modal attention decoder adopts an LSTM-based decoder that contains semantic and acoustic attention modules. Experiments demonstrate that our proposed model achieves state-of-the-art performance on the Clotho dataset. Our code can be found at https://github.com/WangHelin1997/DCASE2021_Task6_PKU

A Mutual learning framework for Few-shot Sound Event Detection

Although prototypical network (ProtoNet) has proved to be an effective method for few-shot sound event detection, two problems still exist. Firstly, the small-scaled support set is insufficient so that the class prototypes may not represent the class center accurately. Secondly, the feature extractor is task-agnostic (or class-agnostic): the feature extractor is trained with base-class data and directly applied to unseen-class data. To address these issues, we present a novel mutual learning framework with transductive learning, which aims at iteratively updating the class prototypes and feature extractor. More specifically, we propose to update class prototypes with transductive inference to make the class prototypes as close to the true class center as possible. To make the feature extractor to be task-specific, we propose to use the updated class prototypes to fine-tune the feature extractor. After that, a fine-tuned feature extractor further helps produce better class prototypes. Our method achieves the F-score of 38.4$\%$ on the DCASE 2021 Task 5 evaluation set, which won the first place in the few-shot bioacoustic event detection task of Detection and Classification of Acoustic Scenes and Events (DCASE) 2021 Challenge.

Towards Joint Intent Detection and Slot Filling via Higher-order Attention

Intent detection (ID) and Slot filling (SF) are two major tasks in spoken language understanding (SLU). Recently, attention mechanism has been shown to be effective in jointly optimizing these two tasks in an interactive manner. However, latest attention-based works concentrated only on the first-order attention design, while ignoring the exploration of higher-order attention mechanisms. In this paper, we propose a BiLinear attention block, which leverages bilinear pooling to simultaneously exploit both the contextual and channel-wise bilinear attention distributions to capture the second-order interactions between the input intent or slot features. Higher and even infinity order interactions are built by stacking numerous blocks and assigning Exponential Linear Unit (ELU) to blocks. Before the decoding stage, we introduce the Dynamic Feature Fusion Layer to implicitly fuse intent and slot information in a more effective way. Technically, instead of simply concatenating intent and slot features, we first compute two correlation matrices to weight on two features. Furthermore, we present Higher-order Attention Network for the SLU tasks. Experiments on two benchmark datasets show that our approach yields improvements compared with the state-of-the-art approach. We also provide discussion to demonstrate the effectiveness of the proposed approach.

On Pursuit of Designing Multi-modal Transformer for Video Grounding

Video grounding aims to localize the temporal segment corresponding to a sentence query from an untrimmed video. Almost all existing video grounding methods fall into two frameworks: 1) Top-down model: It predefines a set of segment candidates and then conducts segment classification and regression. 2) Bottom-up model: It directly predicts frame-wise probabilities of the referential segment boundaries. However, all these methods are not end-to-end, \ie, they always rely on some time-consuming post-processing steps to refine predictions. To this end, we reformulate video grounding as a set prediction task and propose a novel end-to-end multi-modal Transformer model, dubbed as \textbf{GTR}. Specifically, GTR has two encoders for video and language encoding, and a cross-modal decoder for grounding prediction. To facilitate the end-to-end training, we use a Cubic Embedding layer to transform the raw videos into a set of visual tokens. To better fuse these two modalities in the decoder, we design a new Multi-head Cross-Modal Attention. The whole GTR is optimized via a Many-to-One matching loss. Furthermore, we conduct comprehensive studies to investigate different model design choices. Extensive results on three benchmarks have validated the superiority of GTR. All three typical GTR variants achieve record-breaking performance on all datasets and metrics, with several times faster inference speed.

Self-supervised Contrastive Cross-Modality Representation Learning for Spoken Question Answering

Spoken question answering (SQA) requires fine-grained understanding of both spoken documents and questions for the optimal answer prediction. In this paper, we propose novel training schemes for spoken question answering with a self-supervised training stage and a contrastive representation learning stage. In the self-supervised stage, we propose three auxiliary self-supervised tasks, including utterance restoration, utterance insertion, and question discrimination, and jointly train the model to capture consistency and coherence among speech documents without any additional data or annotations. We then propose to learn noise-invariant utterance representations in a contrastive objective by adopting multiple augmentation strategies, including span deletion and span substitution. Besides, we design a Temporal-Alignment attention to semantically align the speech-text clues in the learned common space and benefit the SQA tasks. By this means, the training schemes can more effectively guide the generation model to predict more proper answers. Experimental results show that our model achieves state-of-the-art results on three SQA benchmarks.