Search

In the study of auditory attention, it has been revealed that there exists a robust correlation between attended speech and elicited neural responses, measurable through electroencephalography (EEG). Therefore, it is possible to use the attention information available within EEG signals to guide the extraction of the target speaker in a cocktail party computationally. In this paper, we present a neuro-guided speaker extraction model, i.e. NeuroSpex, using the EEG response of the listener as the sole auxiliary reference cue to extract attended speech from monaural speech mixtures. We propose a novel EEG signal encoder that captures the attention information. Additionally, we propose a cross-attention (CA) mechanism to enhance the speech feature representations, generating a speaker extraction mask. Experimental results on a publicly available dataset demonstrate that our proposed model outperforms two baseline models across various evaluation metrics.

Humans naturally perform audiovisual speech recognition (AVSR), enhancing the accuracy and robustness by integrating auditory and visual information. Spiking neural networks (SNNs), which mimic the brain's information-processing mechanisms, are well-suited for emulating the human capability of AVSR. Despite their potential, research on SNNs for AVSR is scarce, with most existing audio-visual multimodal methods focused on object or digit recognition. These models simply integrate features from both modalities, neglecting their unique characteristics and interactions. Additionally, they often rely on future information for current processing, which increases recognition latency and limits real-time applicability. Inspired by human speech perception, this paper proposes a novel human-inspired SNN named HI-AVSNN for AVSR, incorporating three key characteristics: cueing interaction, causal processing and spike activity. For cueing interaction, we propose a visual-cued auditory attention module (VCA2M) that leverages visual cues to guide attention to auditory features. We achieve causal processing by aligning the SNN's temporal dimension with that of visual and auditory features and applying temporal masking to utilize only past and current information. To implement spike activity, in addition to using SNNs, we leverage the event camera to capture lip movement as spikes, mimicking the human retina and providing efficient visual data. We evaluate HI-AVSNN on an audiovisual speech recognition dataset combining the DVS-Lip dataset with its corresponding audio samples. Experimental results demonstrate the superiority of our proposed fusion method, outperforming existing audio-visual SNN fusion methods and achieving a 2.27% improvement in accuracy over the only existing SNN-based AVSR method.

Conversational Speech Synthesis (CSS) aims to express a target utterance with the proper speaking style in a user-agent conversation setting. Existing CSS methods employ effective multi-modal context modeling techniques to achieve empathy understanding and expression. However, they often need to design complex network architectures and meticulously optimize the modules within them. In addition, due to the limitations of small-scale datasets containing scripted recording styles, they often fail to simulate real natural conversational styles. To address the above issues, we propose a novel generative expressive CSS system, termed GPT-Talker.We transform the multimodal information of the multi-turn dialogue history into discrete token sequences and seamlessly integrate them to form a comprehensive user-agent dialogue context. Leveraging the power of GPT, we predict the token sequence, that includes both semantic and style knowledge, of response for the agent. After that, the expressive conversational speech is synthesized by the conversation-enriched VITS to deliver feedback to the user.Furthermore, we propose a large-scale Natural CSS Dataset called NCSSD, that includes both naturally recorded conversational speech in improvised styles and dialogues extracted from TV shows. It encompasses both Chinese and English languages, with a total duration of 236 hours.We conducted comprehensive experiments on the reliability of the NCSSD and the effectiveness of our GPT-Talker. Both subjective and objective evaluations demonstrate that our model outperforms other state-of-the-art CSS systems significantly in terms of naturalness and expressiveness. The Code, Dataset, and Pre-trained Model are available at: https://github.com/AI-S2-Lab/GPT-Talker., Comment: 14 pages, 6 figures, 8 tables. Accepted by ACM MM 2024

The human brain has the capability to associate the unknown person's voice and face by leveraging their general relationship, referred to as ``cross-modal speaker verification''. This task poses significant challenges due to the complex relationship between the modalities. In this paper, we propose a ``Multi-stage Face-voice Association Learning with Keynote Speaker Diarization''~(MFV-KSD) framework. MFV-KSD contains a keynote speaker diarization front-end to effectively address the noisy speech inputs issue. To balance and enhance the intra-modal feature learning and inter-modal correlation understanding, MFV-KSD utilizes a novel three-stage training strategy. Our experimental results demonstrated robust performance, achieving the first rank in the 2024 Face-voice Association in Multilingual Environments (FAME) challenge with an overall Equal Error Rate (EER) of 19.9%. Details can be found in https://github.com/TaoRuijie/MFV-KSD.

Speaker individuality information is among the most critical elements within speech signals. By thoroughly and accurately modeling this information, it can be utilized in various intelligent speech applications, such as speaker recognition, speaker diarization, speech synthesis, and target speaker extraction. In this article, we aim to present, from a unique perspective, the developmental history, paradigm shifts, and application domains of speaker modeling technologies within the context of deep representation learning framework. This review is designed to provide a clear reference for researchers in the speaker modeling field, as well as for those who wish to apply speaker modeling techniques to specific downstream tasks.

Amid the burgeoning development of generative models like diffusion models, the task of differentiating synthesized audio from its natural counterpart grows more daunting. Deepfake detection offers a viable solution to combat this challenge. Yet, this defensive measure unintentionally fuels the continued refinement of generative models. Watermarking emerges as a proactive and sustainable tactic, preemptively regulating the creation and dissemination of synthesized content. Thus, this paper, as a pioneer, proposes the generative robust audio watermarking method (Groot), presenting a paradigm for proactively supervising the synthesized audio and its source diffusion models. In this paradigm, the processes of watermark generation and audio synthesis occur simultaneously, facilitated by parameter-fixed diffusion models equipped with a dedicated encoder. The watermark embedded within the audio can subsequently be retrieved by a lightweight decoder. The experimental results highlight Groot's outstanding performance, particularly in terms of robustness, surpassing that of the leading state-of-the-art methods. Beyond its impressive resilience against individual post-processing attacks, Groot exhibits exceptional robustness when facing compound attacks, maintaining an average watermark extraction accuracy of around 95%., Comment: Accepted by ACM MM 2024

It was shown that pre-trained models with self-supervised learning (SSL) techniques are effective in various downstream speech tasks. However, most such models are trained on single-speaker speech data, limiting their effectiveness in mixture speech. This motivates us to explore pre-training on mixture speech. This work presents SA-WavLM, a novel pre-trained model for mixture speech. Specifically, SA-WavLM follows an "extract-merge-predict" pipeline in which the representations of each speaker in the input mixture are first extracted individually and then merged before the final prediction. In this pipeline, SA-WavLM performs speaker-informed extractions with the consideration of the interactions between different speakers. Furthermore, a speaker shuffling strategy is proposed to enhance the robustness towards the speaker absence. Experiments show that SA-WavLM either matches or improves upon the state-of-the-art pre-trained models., Comment: InterSpeech 2024

Emotion and Intent Joint Understanding in Multimodal Conversation (MC-EIU) aims to decode the semantic information manifested in a multimodal conversational history, while inferring the emotions and intents simultaneously for the current utterance. MC-EIU is enabling technology for many human-computer interfaces. However, there is a lack of available datasets in terms of annotation, modality, language diversity, and accessibility. In this work, we propose an MC-EIU dataset, which features 7 emotion categories, 9 intent categories, 3 modalities, i.e., textual, acoustic, and visual content, and two languages, i.e., English and Mandarin. Furthermore, it is completely open-source for free access. To our knowledge, MC-EIU is the first comprehensive and rich emotion and intent joint understanding dataset for multimodal conversation. Together with the release of the dataset, we also develop an Emotion and Intent Interaction (EI$^2$) network as a reference system by modeling the deep correlation between emotion and intent in the multimodal conversation. With comparative experiments and ablation studies, we demonstrate the effectiveness of the proposed EI$^2$ method on the MC-EIU dataset. The dataset and codes will be made available at: https://github.com/MC-EIU/MC-EIU., Comment: 26 pages, 8 figures, 12 tables, NeurIPS 2024 Dataset and Benchmark Track

Large language models (LLMs) have demonstrated emergent capabilities across diverse reasoning tasks via popular Chains-of-Thought (COT) prompting. However, such a simple and fast COT approach often encounters limitations in dealing with complicated problems, while a thorough method, which considers multiple reasoning pathways and verifies each step carefully, results in slower inference. This paper addresses the challenge of enabling LLMs to autonomously select between fast and slow inference methods, thereby optimizing both efficiency and effectiveness. We introduce a dynamic decision-making framework that categorizes tasks into two distinct pathways: 'Fast', designated for tasks where the LLM quickly identifies a high-confidence solution, and 'Slow', allocated for tasks that the LLM perceives as complex and for which it has low confidence in immediate solutions as well as requiring more reasoning paths to verify. Experiments on five popular reasoning benchmarks demonstrated the superiority of the DynaThink over baselines.

This paper proposes RefXVC, a method for cross-lingual voice conversion (XVC) that leverages reference information to improve conversion performance. Previous XVC works generally take an average speaker embedding to condition the speaker identity, which does not account for the changing timbre of speech that occurs with different pronunciations. To address this, our method uses both global and local speaker embeddings to capture the timbre changes during speech conversion. Additionally, we observed a connection between timbre and pronunciation in different languages and utilized this by incorporating a timbre encoder and a pronunciation matching network into our model. Furthermore, we found that the variation in tones is not adequately reflected in a sentence, and therefore, we used multiple references to better capture the range of a speaker's voice. The proposed method outperformed existing systems in terms of both speech quality and speaker similarity, highlighting the effectiveness of leveraging reference information in cross-lingual voice conversion. The converted speech samples can be found on the website: \url{http://refxvc.dn3point.com}, Comment: Manuscript under review by TASLP

We introduce a novel multimodality synergistic knowledge distillation scheme tailored for efficient single-eye motion recognition tasks. This method allows a lightweight, unimodal student spiking neural network (SNN) to extract rich knowledge from an event-frame multimodal teacher network. The core strength of this approach is its ability to utilize the ample, coarser temporal cues found in conventional frames for effective emotion recognition. Consequently, our method adeptly interprets both temporal and spatial information from the conventional frame domain, eliminating the need for specialized sensing devices, e.g., event-based camera. The effectiveness of our approach is thoroughly demonstrated using both existing and our compiled single-eye emotion recognition datasets, achieving unparalleled performance in accuracy and efficiency over existing state-of-the-art methods., Comment: Accepted by IJCAI 2024

Data selection for fine-tuning Large Language Models (LLMs) aims to select a high-quality subset from a given candidate dataset to train a Pending Fine-tune Model (PFM) into a Selective-Enhanced Model (SEM). It can improve the model performance and accelerate the training process. Although a few surveys have investigated related works of data selection, there is a lack of comprehensive comparison between existing methods due to their various experimental settings. To address this issue, we first propose a three-stage scheme for data selection and comprehensively review existing works according to this scheme. Then, we design a unified comparing method with ratio-based efficiency indicators and ranking-based feasibility indicators to overcome the difficulty of comparing various models with diverse experimental settings. After an in-depth comparative analysis, we find that the more targeted method with data-specific and model-specific quality labels has higher efficiency, but the introduction of additional noise information should be avoided when designing selection algorithms. Finally, we summarize the trends in data selection and highlight the short-term and long-term challenges to guide future research.

Speech encompasses a wealth of information, including but not limited to content, paralinguistic, and environmental information. This comprehensive nature of speech significantly impacts communication and is crucial for human-computer interaction. Chat-Oriented Large Language Models (LLMs), known for their general-purpose assistance capabilities, have evolved to handle multi-modal inputs, including speech. Although these models can be adept at recognizing and analyzing speech, they often fall short of generating appropriate responses. We argue that this is due to the lack of principles on task definition and model development, which requires open-source datasets and metrics suitable for model evaluation. To bridge the gap, we present SD-Eval, a benchmark dataset aimed at multidimensional evaluation of spoken dialogue understanding and generation. SD-Eval focuses on paralinguistic and environmental information and includes 7,303 utterances, amounting to 8.76 hours of speech data. The data is aggregated from eight public datasets, representing four perspectives: emotion, accent, age, and background sound. To assess the SD-Eval benchmark dataset, we implement three different models and construct a training set following a similar process as SD-Eval. The training set contains 1,052.72 hours of speech data and 724.4k utterances. We also conduct a comprehensive evaluation using objective evaluation methods (e.g. BLEU and ROUGE), subjective evaluations and LLM-based metrics for the generated responses. Models conditioned with paralinguistic and environmental information outperform their counterparts in both objective and subjective measures. Moreover, experiments demonstrate LLM-based metrics show a higher correlation with human evaluation compared to traditional metrics. We open-source SD-Eval at https://github.com/amphionspace/SD-Eval.

The use of Transformer architectures has facilitated remarkable progress in speech enhancement. Training Transformers using substantially long speech utterances is often infeasible as self-attention suffers from quadratic complexity. It is a critical and unexplored challenge for a Transformer-based speech enhancement model to learn from short speech utterances and generalize to longer ones. In this paper, we conduct comprehensive experiments to explore the length generalization problem in speech enhancement with Transformer. Our findings first establish that position embedding provides an effective instrument to alleviate the impact of utterance length on Transformer-based speech enhancement. Specifically, we explore four different position embedding schemes to enable length generalization. The results confirm the superiority of relative position embeddings (RPEs) over absolute PE (APEs) in length generalization., Comment: Accepted by INTERSPEECH 2024

Synthesizing speech across different accents while preserving the speaker identity is essential for various real-world customer applications. However, the individual and accurate modeling of accents and speakers in a text-to-speech (TTS) system is challenging due to the complexity of accent variations and the intrinsic entanglement between the accent and speaker identity. In this paper, we present a novel approach for multi-speaker multi-accent TTS synthesis, which aims to synthesize voices of multiple speakers, each with various accents. Our proposed approach employs a multi-scale accent modeling strategy to address accent variations at different levels. Specifically, we introduce both global (utterance level) and local (phoneme level) accent modeling, supervised by individual accent classifiers to capture the overall variation within accented utterances and fine-grained variations between phonemes, respectively. To control accents and speakers separately, speaker-independent accent modeling is necessary, which is achieved by adversarial training with speaker classifiers to disentangle speaker identity within the multi-scale accent modeling. Consequently, we obtain speaker-independent and accent-discriminative multi-scale embeddings as comprehensive accent features. Additionally, we propose a local accent prediction model that allows to generate accented speech directly from phoneme inputs. Extensive experiments are conducted on an accented English speech corpus. Both objective and subjective evaluations show the superiority of our proposed system compared to baselines systems. Detailed component analysis demonstrates the effectiveness of global and local accent modeling, and speaker disentanglement on multi-speaker multi-accent speech synthesis.

Keyword Spotting (KWS) is essential in edge computing requiring rapid and energy-efficient responses. Spiking Neural Networks (SNNs) are well-suited for KWS for their efficiency and temporal capacity for speech. To further reduce the latency and energy consumption, this study introduces ED-sKWS, an SNN-based KWS model with an early-decision mechanism that can stop speech processing and output the result before the end of speech utterance. Furthermore, we introduce a Cumulative Temporal (CT) loss that can enhance prediction accuracy at both the intermediate and final timesteps. To evaluate early-decision performance, we present the SC-100 dataset including 100 speech commands with beginning and end timestamp annotation. Experiments on the Google Speech Commands v2 and our SC-100 datasets show that ED-sKWS maintains competitive accuracy with 61% timesteps and 52% energy consumption compared to SNN models without early-decision mechanism, ensuring rapid response and energy efficiency., Comment: Accepted by INTERSPEECH2024

Traditional speaker diarization seeks to detect ``who spoke when'' according to speaker characteristics. Extending to target speech diarization, we detect ``when target event occurs'' according to the semantic characteristics of speech. We propose a novel Multimodal Target Speech Diarization (MM-TSD) framework, which accommodates diverse and multi-modal prompts to specify target events in a flexible and user-friendly manner, including semantic language description, pre-enrolled speech, pre-registered face image, and audio-language logical prompts. We further propose a voice-face aligner module to project human voice and face representation into a shared space. We develop a multi-modal dataset based on VoxCeleb2 for MM-TSD training and evaluation. Additionally, we conduct comparative analysis and ablation studies for each category of prompts to validate the efficacy of each component in the proposed framework. Furthermore, our framework demonstrates versatility in performing various signal processing tasks, including speaker diarization and overlap speech detection, using task-specific prompts. MM-TSD achieves robust and comparable performance as a unified system compared to specialized models. Moreover, MM-TSD shows capability to handle complex conversations for real-world dataset., Comment: 13 pages, 7 figures

Audio language models have recently emerged as a promising approach for various audio generation tasks, relying on audio tokenizers to encode waveforms into sequences of discrete symbols. Audio tokenization often poses a necessary compromise between code bitrate and reconstruction accuracy. When dealing with low-bitrate audio codes, language models are constrained to process only a subset of the information embedded in the audio, which in turn restricts their generative capabilities. To circumvent these issues, we propose encoding audio as vector sequences in continuous space $\mathbb R^d$ and autoregressively generating these sequences using a decoder-only diffusion transformer (ARDiT). Our findings indicate that ARDiT excels in zero-shot text-to-speech and exhibits performance that compares to or even surpasses that of state-of-the-art models. High-bitrate continuous speech representation enables almost flawless reconstruction, allowing our model to achieve nearly perfect speech editing. Our experiments reveal that employing Integral Kullback-Leibler (IKL) divergence for distillation at each autoregressive step significantly boosts the perceived quality of the samples. Simultaneously, it condenses the iterative sampling process of the diffusion model into a single step. Furthermore, ARDiT can be trained to predict several continuous vectors in one step, significantly reducing latency during sampling. Impressively, one of our models can generate $170$ ms of $24$ kHz speech per evaluation step with minimal degradation in performance. Audio samples are available at http://ardit-tts.github.io/ .

Partially manipulating a sentence can greatly change its meaning. Recent work shows that countermeasures (CMs) trained on partially spoofed audio can effectively detect such spoofing. However, the current understanding of the decision-making process of CMs is limited. We utilize Grad-CAM and introduce a quantitative analysis metric to interpret CMs' decisions. We find that CMs prioritize the artifacts of transition regions created when concatenating bona fide and spoofed audio. This focus differs from that of CMs trained on fully spoofed audio, which concentrate on the pattern differences between bona fide and spoofed parts. Our further investigation explains the varying nature of CMs' focus while making correct or incorrect predictions. These insights provide a basis for the design of CM models and the creation of datasets. Moreover, this work lays a foundation of interpretability in the field of partial spoofed audio detection that has not been well explored previously., Comment: Accepted at Interspeech 2024

Mainstream approaches to aligning large language models (LLMs) heavily rely on human preference data, particularly when models require periodic updates. The standard process for iterative alignment of LLMs involves collecting new human feedback for each update. However, the data collection process is costly and challenging to scale. To address this issue, we introduce the "TS-Align" framework, which fine-tunes a policy model using pairwise feedback data automatically mined from its outputs. This automatic mining process is efficiently accomplished through the collaboration between a large-scale teacher model and a small-scale student model. The policy fine-tuning process can be iteratively repeated using on-policy generations within our proposed teacher-student collaborative framework. Through extensive experiments, we demonstrate that our final aligned policy outperforms the base policy model with an average win rate of 69.7% across seven conversational or instruction-following datasets. Furthermore, we show that the ranking capability of the teacher is effectively distilled into the student through our pipeline, resulting in a small-scale yet effective reward model for policy model alignment.

The advancement of large language models (LLMs) has propelled the development of dialogue systems. Unlike the popular ChatGPT-like assistant model, which only satisfies the user's preferences, task-oriented dialogue systems have also faced new requirements and challenges in the broader business field. They are expected to provide correct responses at each dialogue turn, at the same time, achieve the overall goal defined by the task. By understanding rhetorical structures and topic structures via topic segmentation and discourse parsing, a dialogue system may do a better planning to achieve both objectives. However, while both structures belong to discourse structure in linguistics, rhetorical structure and topic structure are mostly modeled separately or with one assisting the other in the prior work. The interaction between these two structures has not been considered for joint modeling and mutual learning. Furthermore, unsupervised learning techniques to achieve the above are not well explored. To fill this gap, we propose an unsupervised mutual learning framework of two structures leveraging the global and local connections between them. We extend the topic modeling between non-adjacent discourse units to ensure global structural relevance with rhetorical structures. We also incorporate rhetorical structures into the topic structure through a graph neural network model to ensure local coherence consistency. Finally, we utilize the similarity between the two fused structures for mutual learning. The experimental results demonstrate that our methods outperform all strong baselines on two dialogue rhetorical datasets (STAC and Molweni), as well as dialogue topic datasets (Doc2Dial and TIAGE). We provide our code at https://github.com/Jeff-Sue/URT.

The automatic evaluation of natural language generation (NLG) systems presents a long-lasting challenge. Recent studies have highlighted various neural metrics that align well with human evaluations. Yet, the robustness of these evaluators against adversarial perturbations remains largely under-explored due to the unique challenges in obtaining adversarial data for different NLG evaluation tasks. To address the problem, we introduce AdvEval, a novel black-box adversarial framework against NLG evaluators. AdvEval is specially tailored to generate data that yield strong disagreements between human and victim evaluators. Specifically, inspired by the recent success of large language models (LLMs) in text generation and evaluation, we adopt strong LLMs as both the data generator and gold evaluator. Adversarial data are automatically optimized with feedback from the gold and victim evaluator. We conduct experiments on 12 victim evaluators and 11 NLG datasets, spanning tasks including dialogue, summarization, and question evaluation. The results show that AdvEval can lead to significant performance degradation of various victim metrics, thereby validating its efficacy., Comment: ACL24 Finding

Transformer and its derivatives have achieved success in diverse tasks across computer vision, natural language processing, and speech processing. To reduce the complexity of computations within the multi-head self-attention mechanism in Transformer, Selective State Space Models (i.e., Mamba) were proposed as an alternative. Mamba exhibited its effectiveness in natural language processing and computer vision tasks, but its superiority has rarely been investigated in speech signal processing. This paper explores solutions for applying Mamba to speech processing using two typical speech processing tasks: speech recognition, which requires semantic and sequential information, and speech enhancement, which focuses primarily on sequential patterns. The experimental results exhibit the superiority of bidirectional Mamba (BiMamba) for speech processing to vanilla Mamba. Moreover, experiments demonstrate the effectiveness of BiMamba as an alternative to the self-attention module in Transformer and its derivates, particularly for the semantic-aware task. The crucial technologies for transferring Mamba to speech are then summarized in ablation studies and the discussion section to offer insights for future research.

It remains a challenge to effectively control the emotion rendering in text-to-speech (TTS) synthesis. Prior studies have primarily focused on learning a global prosodic representation at the utterance level, which strongly correlates with linguistic prosody. Our goal is to construct a hierarchical emotion distribution (ED) that effectively encapsulates intensity variations of emotions at various levels of granularity, encompassing phonemes, words, and utterances. During TTS training, the hierarchical ED is extracted from the ground-truth audio and guides the predictor to establish a connection between emotional and linguistic prosody. At run-time inference, the TTS model generates emotional speech and, at the same time, provides quantitative control of emotion over the speech constituents. Both objective and subjective evaluations validate the effectiveness of the proposed framework in terms of emotion prediction and control., Comment: This is accepted to IEEE ICASSP 2024

This paper aims to efficiently enable large language models (LLMs) to use external knowledge and goal guidance in conversational recommender system (CRS) tasks. Advanced LLMs (e.g., ChatGPT) are limited in domain-specific CRS tasks for 1) generating grounded responses with recommendation-oriented knowledge, or 2) proactively leading the conversations through different dialogue goals. In this work, we first analyze those limitations through a comprehensive evaluation, showing the necessity of external knowledge and goal guidance which contribute significantly to the recommendation accuracy and language quality. In light of this finding, we propose a novel ChatCRS framework to decompose the complex CRS task into several sub-tasks through the implementation of 1) a knowledge retrieval agent using a tool-augmented approach to reason over external Knowledge Bases and 2) a goal-planning agent for dialogue goal prediction. Experimental results on two multi-goal CRS datasets reveal that ChatCRS sets new state-of-the-art benchmarks, improving language quality of informativeness by 17% and proactivity by 27%, and achieving a tenfold enhancement in recommendation accuracy., Comment: Main paper 8 pages; References and Appendix 9 pages; 7 figures and 14 tables

Audio-visual target speaker extraction (AV-TSE) aims to extract the specific person's speech from the audio mixture given auxiliary visual cues. Previous methods usually search for the target voice through speech-lip synchronization. However, this strategy mainly focuses on the existence of target speech, while ignoring the variations of the noise characteristics. That may result in extracting noisy signals from the incorrect sound source in challenging acoustic situations. To this end, we propose a novel reverse selective auditory attention mechanism, which can suppress interference speakers and non-speech signals to avoid incorrect speaker extraction. By estimating and utilizing the undesired noisy signal through this mechanism, we design an AV-TSE framework named Subtraction-and-ExtrAction network (SEANet) to suppress the noisy signals. We conduct abundant experiments by re-implementing three popular AV-TSE methods as the baselines and involving nine metrics for evaluation. The experimental results show that our proposed SEANet achieves state-of-the-art results and performs well for all five datasets. We will release the codes, the models and data logs.

Generative Adversarial Network (GAN) based vocoders are superior in both inference speed and synthesis quality when reconstructing an audible waveform from an acoustic representation. This study focuses on improving the discriminator for GAN-based vocoders. Most existing Time-Frequency Representation (TFR)-based discriminators are rooted in Short-Time Fourier Transform (STFT), which owns a constant Time-Frequency (TF) resolution, linearly scaled center frequencies, and a fixed decomposition basis, making it incompatible with signals like singing voices that require dynamic attention for different frequency bands and different time intervals. Motivated by that, we propose a Multi-Scale Sub-Band Constant-Q Transform CQT (MS-SB-CQT) discriminator and a Multi-Scale Temporal-Compressed Continuous Wavelet Transform CWT (MS-TC-CWT) discriminator. Both CQT and CWT have a dynamic TF resolution for different frequency bands. In contrast, CQT has a better modeling ability in pitch information, and CWT has a better modeling ability in short-time transients. Experiments conducted on both speech and singing voices confirm the effectiveness of our proposed discriminators. Moreover, the STFT, CQT, and CWT-based discriminators can be used jointly for better performance. The proposed discriminators can boost the synthesis quality of various state-of-the-art GAN-based vocoders, including HiFi-GAN, BigVGAN, and APNet., Comment: arXiv admin note: text overlap with arXiv:2311.14957

Modern speaker recognition system relies on abundant and balanced datasets for classification training. However, diverse defective datasets, such as partially-labelled, small-scale, and imbalanced datasets, are common in real-world applications. Previous works usually studied specific solutions for each scenario from the algorithm perspective. However, the root cause of these problems lies in dataset imperfections. To address these challenges with a unified solution, we propose the Voice Conversion Augmentation (VCA) strategy to obtain pseudo speech from the training set. Furthermore, to guarantee generation quality, we designed the VCA-NN~(nearest neighbours) strategy to select source speech from utterances that are close to the target speech in the representation space. Our experimental results on three created datasets demonstrated that VCA-NN effectively mitigates these dataset problems, which provides a new direction for handling the speaker recognition problems from the data aspect., Comment: 5 pages

Audio-visual active speaker detection (AV-ASD) aims to identify which visible face is speaking in a scene with one or more persons. Most existing AV-ASD methods prioritize capturing speech-lip correspondence. However, there is a noticeable gap in addressing the challenges from real-world AV-ASD scenarios. Due to the presence of low-quality noisy videos in such cases, AV-ASD systems without a selective listening ability are short of effectively filtering out disruptive voice components from mixed audio inputs. In this paper, we propose a Multi-modal Speaker Extraction-to-Detection framework named `MuSED', which is pre-trained with audio-visual target speaker extraction to learn the denoising ability, then it is fine-tuned with the AV-ASD task. Meanwhile, to better capture the multi-modal information and deal with real-world problems such as missing modality, MuSED is modelled on the time domain directly and integrates the multi-modal plus-and-minus augmentation strategy. Our experiments demonstrate that MuSED substantially outperforms the state-of-the-art AV-ASD methods and achieves 95.6% mAP on the AVA-ActiveSpeaker dataset, 98.3% AP on the ASW dataset, and 97.9% F1 on the Columbia AV-ASD dataset, respectively. We will publicly release the code in due course., Comment: 10 pages

Audio-visual target speech extraction (AV-TSE) is one of the enabling technologies in robotics and many audio-visual applications. One of the challenges of AV-TSE is how to effectively utilize audio-visual synchronization information in the process. AV-HuBERT can be a useful pre-trained model for lip-reading, which has not been adopted by AV-TSE. In this paper, we would like to explore the way to integrate a pre-trained AV-HuBERT into our AV-TSE system. We have good reasons to expect an improved performance. To benefit from the inter and intra-modality correlations, we also propose a novel Mask-And-Recover (MAR) strategy for self-supervised learning. The experimental results on the VoxCeleb2 dataset show that our proposed model outperforms the baselines both in terms of subjective and objective metrics, suggesting that the pre-trained AV-HuBERT model provides more informative visual cues for target speech extraction. Furthermore, through a comparative study, we confirm that the proposed Mask-And-Recover strategy is significantly effective., Comment: Accepted by IJCNN 2024

Training or finetuning large-scale language models (LLMs) requires substantial computation resources, motivating recent efforts to explore parameter-efficient adaptation to downstream tasks. One approach is to treat these models as black boxes and use forward passes (Inference APIs) to interact with them. Current research focuses on adapting these black-box models to downstream tasks using gradient-free prompt optimization, but this often involves an expensive process of searching task-specific prompts. Therefore, we are motivated to study black-box language model adaptation without prompt search. Specifically, we introduce a label-enhanced cross-attention network called CrossTune, which models the semantic relatedness between the input text sequence and task-specific label descriptions. Its effectiveness is examined in the context of few-shot text classification. To improve the generalization of CrossTune, we utilize ChatGPT to generate additional training data through in-context learning. A switch mechanism is implemented to exclude low-quality ChatGPT-generated data. Through extensive experiments on seven benchmark text classification datasets, we demonstrate that our proposed approach outperforms the previous state-of-the-art gradient-free black-box tuning method by 5.7% on average. Even without using ChatGPT-augmented data, CrossTune performs better or comparably than previous black-box tuning methods, suggesting the effectiveness of our approach., Comment: Accepted by LREC-Coling 2024

Speech applications are expected to be low-power and robust under noisy conditions. An effective Voice Activity Detection (VAD) front-end lowers the computational need. Spiking Neural Networks (SNNs) are known to be biologically plausible and power-efficient. However, SNN-based VADs have yet to achieve noise robustness and often require large models for high performance. This paper introduces a novel SNN-based VAD model, referred to as sVAD, which features an auditory encoder with an SNN-based attention mechanism. Particularly, it provides effective auditory feature representation through SincNet and 1D convolution, and improves noise robustness with attention mechanisms. The classifier utilizes Spiking Recurrent Neural Networks (sRNN) to exploit temporal speech information. Experimental results demonstrate that our sVAD achieves remarkable noise robustness and meanwhile maintains low power consumption and a small footprint, making it a promising solution for real-world VAD applications., Comment: Accepted by ICASSP 2024

Despite the vast repository of global medical knowledge predominantly being in English, local languages are crucial for delivering tailored healthcare services, particularly in areas with limited medical resources. To extend the reach of medical AI advancements to a broader population, we aim to develop medical LLMs across the six most widely spoken languages, encompassing a global population of 6.1 billion. This effort culminates in the creation of the ApolloCorpora multilingual medical dataset and the XMedBench benchmark. In the multilingual medical benchmark, the released Apollo models, at various relatively-small sizes (i.e., 0.5B, 1.8B, 2B, 6B, and 7B), achieve the best performance among models of equivalent size. Especially, Apollo-7B is the state-of-the-art multilingual medical LLMs up to 70B. Additionally, these lite models could be used to improve the multi-lingual medical capabilities of larger models without fine-tuning in a proxy-tuning fashion. We will open-source training corpora, code, model weights and evaluation benchmark., Comment: Preprint

It remains a significant challenge how to quantitatively control the expressiveness of speech emotion in speech generation. In this work, we present a novel approach for manipulating the rendering of emotions for speech generation. We propose a hierarchical emotion distribution extractor, i.e. Hierarchical ED, that quantifies the intensity of emotions at different levels of granularity. Support vector machines (SVMs) are employed to rank emotion intensity, resulting in a hierarchical emotional embedding. Hierarchical ED is subsequently integrated into the FastSpeech2 framework, guiding the model to learn emotion intensity at phoneme, word, and utterance levels. During synthesis, users can manually edit the emotional intensity of the generated voices. Both objective and subjective evaluations demonstrate the effectiveness of the proposed network in terms of fine-grained quantitative emotion editing., Comment: This paper is submitted to IEEE Signal Processing Letters

Brain-inspired spiking neural networks (SNNs) have gained prominence in the field of neuromorphic computing owing to their low energy consumption during feedforward inference on neuromorphic hardware. However, it remains an open challenge how to effectively benefit from the sparse event-driven property of SNNs to minimize backpropagation learning costs. In this paper, we conduct a comprehensive examination of the existing event-driven learning algorithms, reveal their limitations, and propose novel solutions to overcome them. Specifically, we introduce two novel event-driven learning methods: the spike-timing-dependent event-driven (STD-ED) and membrane-potential-dependent event-driven (MPD-ED) algorithms. These proposed algorithms leverage precise neuronal spike timing and membrane potential, respectively, for effective learning. The two methods are extensively evaluated on static and neuromorphic datasets to confirm their superior performance. They outperform existing event-driven counterparts by up to 2.51% for STD-ED and 6.79% for MPD-ED on the CIFAR-100 dataset. In addition, we theoretically and experimentally validate the energy efficiency of our methods on neuromorphic hardware. On-chip learning experiments achieved a remarkable 30-fold reduction in energy consumption over time-step-based surrogate gradient methods. The demonstrated efficiency and efficacy of the proposed event-driven learning methods emphasize their potential to significantly advance the fields of neuromorphic computing, offering promising avenues for energy-efficiency applications.

Human language can be expressed in either written or spoken form, i.e. text or speech. Humans can acquire knowledge from text to improve speaking and listening. However, the quest for speech pre-trained models to leverage unpaired text has just started. In this paper, we investigate a new way to pre-train such a joint speech-text model to learn enhanced speech representations and benefit various speech-related downstream tasks. Specifically, we propose a novel pre-training method, text-guided HuBERT, or T-HuBERT, which performs self-supervised learning over speech to derive phoneme-like discrete representations. And these phoneme-like pseudo-label sequences are firstly derived from speech via the generative adversarial networks (GAN) to be statistically similar to those from additional unpaired textual data. In this way, we build a bridge between unpaired speech and text in an unsupervised manner. Extensive experiments demonstrate the significant superiority of our proposed method over various strong baselines, which achieves up to 15.3% relative Word Error Rate (WER) reduction on the LibriSpeech dataset., Comment: 5 pages, 1 figures,5 tables, accepted by IEEE Signal Processing Letters(SPL)

Cued Speech (CS) is a pure visual coding method used by hearing-impaired people that combines lip reading with several specific hand shapes to make the spoken language visible. Automatic CS recognition (ACSR) seeks to transcribe visual cues of speech into text, which can help hearing-impaired people to communicate effectively. The visual information of CS contains lip reading and hand cueing, thus the fusion of them plays an important role in ACSR. However, most previous fusion methods struggle to capture the global dependency present in long sequence inputs of multi-modal CS data. As a result, these methods generally fail to learn the effective cross-modal relationships that contribute to the fusion. Recently, attention-based transformers have been a prevalent idea for capturing the global dependency over the long sequence in multi-modal fusion, but existing multi-modal fusion transformers suffer from both poor recognition accuracy and inefficient computation for the ACSR task. To address these problems, we develop a novel computation and parameter efficient multi-modal fusion transformer by proposing a novel Token-Importance-Aware Attention mechanism (TIAA), where a token utilization rate (TUR) is formulated to select the important tokens from the multi-modal streams. More precisely, TIAA firstly models the modality-specific fine-grained temporal dependencies over all tokens of each modality, and then learns the efficient cross-modal interaction for the modality-shared coarse-grained temporal dependencies over the important tokens of different modalities. Besides, a light-weight gated hidden projection is designed to control the feature flows of TIAA. The resulting model, named Economical Cued Speech Fusion Transformer (EcoCued), achieves state-of-the-art performance on all existing CS datasets, compared with existing transformer-based fusion methods and ACSR fusion methods., Comment: Accepted by TASLP

Spiking Neural Networks (SNNs) mimic the information-processing mechanisms of the human brain and are highly energy-efficient, making them well-suited for low-power edge devices. However, the pursuit of accuracy in current studies leads to large, long-timestep SNNs, conflicting with the resource constraints of these devices. In order to design lightweight and efficient SNNs, we propose a new approach named LitE-SNN that incorporates both spatial and temporal compression into the automated network design process. Spatially, we present a novel Compressive Convolution block (CompConv) to expand the search space to support pruning and mixed-precision quantization. Temporally, we are the first to propose a compressive timestep search to identify the optimal number of timesteps under specific computation cost constraints. Finally, we formulate a joint optimization to simultaneously learn the architecture parameters and spatial-temporal compression strategies to achieve high performance while minimizing memory and computation costs. Experimental results on CIFAR-10, CIFAR-100, and Google Speech Command datasets demonstrate our proposed LitE-SNNs can achieve competitive or even higher accuracy with remarkably smaller model sizes and fewer computation costs.

There has been a long-standing quest for a unified audio-visual-text model to enable various multimodal understanding tasks, which mimics the listening, seeing and reading process of human beings. Humans tends to represent knowledge using two separate systems: one for representing verbal (textual) information and one for representing non-verbal (visual and auditory) information. These two systems can operate independently but can also interact with each other. Motivated by this understanding of human cognition, in this paper, we introduce CoAVT -- a novel cognition-inspired Correlated Audio-Visual-Text pre-training model to connect the three modalities. It contains a joint audio-visual encoder that learns to encode audio-visual synchronization information together with the audio and visual content for non-verbal information, and a text encoder to handle textual input for verbal information. To bridge the gap between modalities, CoAVT employs a query encoder, which contains a set of learnable query embeddings, and extracts the most informative audiovisual features of the corresponding text. Additionally, to leverage the correspondences between audio and vision with language respectively, we also establish the audio-text and visual-text bi-modal alignments upon the foundational audiovisual-text tri-modal alignment to enhance the multimodal representation learning. Finally, we jointly optimize CoAVT model with three multimodal objectives: contrastive loss, matching loss and language modeling loss. Extensive experiments show that CoAVT can learn strong multimodal correlations and be generalized to various downstream tasks. CoAVT establishes new state-of-the-art performance on text-video retrieval task on AudioCaps for both zero-shot and fine-tuning settings, audio-visual event classification and audio-visual retrieval tasks on AudioSet and VGGSound.

Transformer architecture has enabled recent progress in speech enhancement. Since Transformers are position-agostic, positional encoding is the de facto standard component used to enable Transformers to distinguish the order of elements in a sequence. However, it remains unclear how positional encoding exactly impacts speech enhancement based on Transformer architectures. In this paper, we perform a comprehensive empirical study evaluating five positional encoding methods, i.e., Sinusoidal and learned absolute position embedding (APE), T5-RPE, KERPLE, as well as the Transformer without positional encoding (No-Pos), across both causal and noncausal configurations. We conduct extensive speech enhancement experiments, involving spectral mapping and masking methods. Our findings establish that positional encoding is not quite helpful for the models in a causal configuration, which indicates that causal attention may implicitly incorporate position information. In a noncausal configuration, the models significantly benefit from the use of positional encoding. In addition, we find that among the four position embeddings, relative position embeddings outperform APEs., Comment: Accepted by ICASSP 2024

Speaker verification is hampered by background noise, particularly at extremely low Signal-to-Noise Ratio (SNR) under 0 dB. It is difficult to suppress noise without introducing unwanted artifacts, which adversely affects speaker verification. We proposed the mechanism called Gradient Weighting (Grad-W), which dynamically identifies and reduces artifact noise during prediction. The mechanism is based on the property that the gradient indicates which parts of the input the model is paying attention to. Specifically, when the speaker network focuses on a region in the denoised utterance but not on the clean counterpart, we consider it artifact noise and assign higher weights for this region during optimization of enhancement. We validate it by training an enhancement model and testing the enhanced utterance on speaker verification. The experimental results show that our approach effectively reduces artifact noise, improving speaker verification across various SNR levels., Comment: Accepted by ICASSP 2024

This paper summarizes our team's efforts in both tracks of the ICMC-ASR Challenge for in-car multi-channel automatic speech recognition. Our submitted systems for ICMC-ASR Challenge include the multi-channel front-end enhancement and diarization, training data augmentation, speech recognition modeling with multi-channel branches. Tested on the offical Eval1 and Eval2 set, our best system achieves a relative 34.3% improvement in CER and 56.5% improvement in cpCER, compared to the offical baseline system., Comment: Technical Report. 2 pages. For ICMC-ASR-2023 Challenge

Automatic evaluation is an integral aspect of dialogue system research. The traditional reference-based NLG metrics are generally found to be unsuitable for dialogue assessment. Consequently, recent studies have suggested various unique, reference-free neural metrics that better align with human evaluations. Notably among them, large language models (LLMs), particularly the instruction-tuned variants like ChatGPT, are shown to be promising substitutes for human judges. Yet, existing works on utilizing LLMs for automatic dialogue evaluation are limited in their scope in terms of the number of meta-evaluation datasets, mode of evaluation, coverage of LLMs, etc. Hence, it remains inconclusive how effective these LLMs are. To this end, we conduct a comprehensive study on the application of LLMs for automatic dialogue evaluation. Specifically, we analyze the multi-dimensional evaluation capability of 30 recently emerged LLMs at both turn and dialogue levels, using a comprehensive set of 12 meta-evaluation datasets. Additionally, we probe the robustness of the LLMs in handling various adversarial perturbations at both turn and dialogue levels. Finally, we explore how model-level and dimension-level ensembles impact the evaluation performance. All resources are available at https://github.com/e0397123/comp-analysis., Comment: An extended version of AAAI-2024 camera-ready paper (appendix included, 16 pages)

Conversational Speech Synthesis (CSS) aims to accurately express an utterance with the appropriate prosody and emotional inflection within a conversational setting. While recognising the significance of CSS task, the prior studies have not thoroughly investigated the emotional expressiveness problems due to the scarcity of emotional conversational datasets and the difficulty of stateful emotion modeling. In this paper, we propose a novel emotional CSS model, termed ECSS, that includes two main components: 1) to enhance emotion understanding, we introduce a heterogeneous graph-based emotional context modeling mechanism, which takes the multi-source dialogue history as input to model the dialogue context and learn the emotion cues from the context; 2) to achieve emotion rendering, we employ a contrastive learning-based emotion renderer module to infer the accurate emotion style for the target utterance. To address the issue of data scarcity, we meticulously create emotional labels in terms of category and intensity, and annotate additional emotional information on the existing conversational dataset (DailyTalk). Both objective and subjective evaluations suggest that our model outperforms the baseline models in understanding and rendering emotions. These evaluations also underscore the importance of comprehensive emotional annotations. Code and audio samples can be found at: https://github.com/walker-hyf/ECSS., Comment: 9 pages, 4 figures, Accepted by AAAI'2024, Code and audio samples: https://github.com/walker-hyf/ECSS

Amphion is an open-source toolkit for Audio, Music, and Speech Generation, targeting to ease the way for junior researchers and engineers into these fields. It presents a unified framework that is inclusive of diverse generation tasks and models, with the added bonus of being easily extendable for new incorporation. The toolkit is designed with beginner-friendly workflows and pre-trained models, allowing both beginners and seasoned researchers to kick-start their projects with relative ease. Additionally, it provides interactive visualizations and demonstrations of classic models for educational purposes. The initial release of Amphion v0.1 supports a range of tasks including Text to Speech (TTS), Text to Audio (TTA), and Singing Voice Conversion (SVC), supplemented by essential components like data preprocessing, state-of-the-art vocoders, and evaluation metrics. This paper presents a high-level overview of Amphion., Comment: Amphion Website: https://github.com/open-mmlab/Amphion

Previous studies demonstrate the impressive performance of residual neural networks (ResNet) in speaker verification. The ResNet models treat the time and frequency dimensions equally. They follow the default stride configuration designed for image recognition, where the horizontal and vertical axes exhibit similarities. This approach ignores the fact that time and frequency are asymmetric in speech representation. In this paper, we address this issue and look for optimal stride configurations specifically tailored for speaker verification. We represent the stride space on a trellis diagram, and conduct a systematic study on the impact of temporal and frequency resolutions on the performance and further identify two optimal points, namely Golden Gemini, which serves as a guiding principle for designing 2D ResNet-based speaker verification models. By following the principle, a state-of-the-art ResNet baseline model gains a significant performance improvement on VoxCeleb, SITW, and CNCeleb datasets with 7.70%/11.76% average EER/minDCF reductions, respectively, across different network depths (ResNet18, 34, 50, and 101), while reducing the number of parameters by 16.5% and FLOPs by 4.1%. We refer to it as Gemini ResNet. Further investigation reveals the efficacy of the proposed Golden Gemini operating points across various training conditions and architectures. Furthermore, we present a new benchmark, namely the Gemini DF-ResNet, using a cutting-edge model., Comment: Accepted to IEEE/ACM Transactions on Audio, Speech, and Language Processing. Open Access: https://ieeexplore.ieee.org/abstract/document/10497864

Learning from human demonstrations is an emerging trend for designing intelligent robotic systems. However, previous methods typically regard videos as instructions, simply dividing them into action sequences for robotic repetition, which poses obstacles to generalization to diverse tasks or object instances. In this paper, we propose a different perspective, considering human demonstration videos not as mere instructions, but as a source of knowledge for robots. Motivated by this perspective and the remarkable comprehension and generalization capabilities exhibited by large language models (LLMs), we propose DigKnow, a method that DIstills Generalizable KNOWledge with a hierarchical structure. Specifically, DigKnow begins by converting human demonstration video frames into observation knowledge. This knowledge is then subjected to analysis to extract human action knowledge and further distilled into pattern knowledge compassing task and object instances, resulting in the acquisition of generalizable knowledge with a hierarchical structure. In settings with different tasks or object instances, DigKnow retrieves relevant knowledge for the current task and object instances. Subsequently, the LLM-based planner conducts planning based on the retrieved knowledge, and the policy executes actions in line with the plan to achieve the designated task. Utilizing the retrieved knowledge, we validate and rectify planning and execution outcomes, resulting in a substantial enhancement of the success rate. Experimental results across a range of tasks and scenes demonstrate the effectiveness of this approach in facilitating real-world robots to accomplish tasks with the knowledge derived from human demonstrations.

Speech enhancement is widely used as a front-end to improve the speech quality in many audio systems, while it is hard to extract the target speech in multi-talker conditions without prior information on the speaker identity. It was shown that the auditory attention on the target speaker can be decoded from the electroencephalogram (EEG) of the listener implicitly. In this work, we therefore propose a novel end-to-end brain-assisted speech enhancement network (BASEN), which incorporates the listeners' EEG signals and adopts a temporal convolutional network together with a convolutional multi-layer cross attention module to fuse EEG-audio features. Considering that an EEG cap with sparse channels exhibits multiple benefits and in practice many electrodes might contribute marginally, we further propose two channel selection methods, called residual Gumbel selection and convolutional regularization selection. They are dedicated to tackling training instability and duplicated channel selections, respectively. Experimental results on a public dataset show the superiority of the proposed BASEN over existing approaches. The proposed channel selection methods can significantly reduce the amount of informative EEG channels with a negligible impact on the performance., Comment: arXiv admin note: text overlap with arXiv:2305.09994

Adapting a language model into a specific domain, a.k.a `domain adaption', is a common practice when specialized knowledge, e.g. medicine, is not encapsulated in a general language model like Llama2. The challenge lies in the heterogeneity of data across the two training stages, as it varies in languages, genres, or formats. To tackle this and simplify the learning protocol, we propose to transform heterogeneous data, from the both pre-training and supervised stages, into a unified, simple input-output pair format. We validate the new protocol in the domains where proprietary LLMs like ChatGPT perform relatively poorly, such as Traditional Chinese Medicine. The developed model, HuatuoGPT-II, has shown state-of-the-art performance in Chinese medicine domain on a number of benchmarks, e.g. medical licensing exams. It even outperforms proprietary models like ChatGPT and GPT-4 in some aspects, especially in Traditional Chinese Medicine. Expert manual evaluations further validate HuatuoGPT-II's advantages over existing LLMs. Notably, HuatuoGPT-II was benchmarked in a fresh Chinese National Medical Licensing Examination where it achieved the best performance, showcasing not only its effectiveness but also its generalization capabilities.

Text embedding models have significantly contributed to advancements in natural language processing by adeptly capturing semantic properties of textual data. However, the ability of these models to generalize across a wide range of syntactic contexts remains under-explored. In this paper, we first develop an evaluation set, named \textbf{SR}, to scrutinize the capability for syntax understanding of text embedding models from two crucial syntactic aspects: Structural heuristics, and Relational understanding among concepts, as revealed by the performance gaps in previous studies. Our findings reveal that existing text embedding models have not sufficiently addressed these syntactic understanding challenges, and such ineffectiveness becomes even more apparent when evaluated against existing benchmark datasets. Furthermore, we conduct rigorous analysis to unearth factors that lead to such limitations and examine why previous evaluations fail to detect such ineffectiveness. Lastly, we propose strategies to augment the generalization ability of text embedding models in diverse syntactic scenarios. This study serves to highlight the hurdles associated with syntactic generalization and provides pragmatic guidance for boosting model performance across varied syntactic contexts., Comment: Accepted to EMNLP-Findings 2023, datasets and code are released