Guangzhi Sun

Siyin Wang, Wenyi Yu, Yudong Yang et al.

Speech quality assessment typically requires evaluating audio from multiple aspects, such as mean opinion score (MOS) and speaker similarity (SIM) \etc., which can be challenging to cover using one small model designed for a single task. In this paper, we propose leveraging recently introduced auditory large language models (LLMs) for automatic speech quality assessment. By employing task-specific prompts, auditory LLMs are finetuned to predict MOS, SIM and A/B testing results, which are commonly used for evaluating text-to-speech systems. Additionally, the finetuned auditory LLM is able to generate natural language descriptions assessing aspects like noisiness, distortion, discontinuity, and overall quality, providing more interpretable outputs. Extensive experiments have been performed on the NISQA, BVCC, SOMOS and VoxSim speech quality datasets, using open-source auditory LLMs such as SALMONN, Qwen-Audio, and Qwen2-Audio. For the natural language descriptions task, a commercial model Google Gemini 1.5 Pro is also evaluated. The results demonstrate that auditory LLMs achieve competitive performance compared to state-of-the-art task-specific small models in predicting MOS and SIM, while also delivering promising results in A/B testing and natural language descriptions. Our data processing scripts and finetuned model checkpoints can be found at https://github.com/bytedance/SALMONN.

Guangzhi Sun, Xianrui Zheng, Chao Zhang et al.

End-to-end automatic speech recognition (ASR) and large language models, such as Whisper and GPT-2, have recently been scaled to use vast amounts of training data. Despite the large amount of training data, infrequent content words that occur in a particular task may still exhibit poor ASR performance, with contextual biasing a possible remedy. This paper investigates the effectiveness of neural contextual biasing for Whisper combined with GPT-2. Specifically, this paper proposes integrating an adapted tree-constrained pointer generator (TCPGen) component for Whisper and a dedicated training scheme to dynamically adjust the final output without modifying any Whisper model parameters. Experiments across three datasets show a considerable reduction in errors on biasing words with a biasing list of 1000 words. Contextual biasing was more effective when applied to domain-specific data and can boost the performance of Whisper and GPT-2 without losing their generality.

Potsawee Manakul, Guangzhi Sun, Warit Sirichotedumrong et al.

Audio language models process audio inputs using textual prompts for tasks like speech recognition and audio captioning. Although built on multilingual pre-trained components, most are trained primarily on English, limiting their usability for other languages. This paper evaluates audio language models on Thai, a low-resource language, and finds that they lack emergent cross-lingual abilities despite their multilingual foundations. To address this, we explore data mixtures that optimize audio language models for both a target language and English while integrating audio comprehension and speech instruction-following into a unified model. Our experiments provide insights into improving instruction-following in low-resource languages by balancing language-specific and multilingual training data. The proposed model, Typhoon-Audio, significantly outperforms existing open-source models and achieves performance comparable to state-of-the-art Gemini-1.5-Pro in both English and Thai.

Yudong Yang, Xuezhen Zhang, Zhifeng Han et al.

Recent progress in large language models (LLMs) has enabled understanding of both speech and non-speech audio, but exposing new safety risks emerging from complex audio inputs that are inadequately handled by current safeguards. We introduce SACRED-Bench (Speech-Audio Composition for RED-teaming) to evaluate the robustness of LLMs under complex audio-based attacks. Unlike existing perturbation-based methods that rely on noise optimization or white-box access, SACRED-Bench exploits speech-audio composition mechanisms. SACRED-Bench adopts three mechanisms: (a) speech overlap and multi-speaker dialogue, which embeds harmful prompts beneath or alongside benign speech; (b) speech-audio mixture, which imply unsafe intent via non-speech audio alongside benign speech or audio; and (c) diverse spoken instruction formats (open-ended QA, yes/no) that evade text-only filters. Experiments show that, even Gemini 2.5 Pro, the state-of-the-art proprietary LLM, still exhibits 66% attack success rate in SACRED-Bench test set, exposing vulnerabilities under cross-modal, speech-audio composition attacks. To bridge this gap, we propose SALMONN-Guard, a safeguard LLM that jointly inspects speech, audio, and text for safety judgments, reducing attack success down to 20%. Our results highlight the need for audio-aware defenses for the safety of multimodal LLMs. The benchmark and SALMONN-Guard checkpoints can be found at https://huggingface.co/datasets/tsinghua-ee/SACRED-Bench. Warning: this paper includes examples that may be offensive or harmful.

Guangzhi Sun, Shutong Feng, Dongcheng Jiang et al.

Recently, advancements in large language models (LLMs) have shown an unprecedented ability across various language tasks. This paper investigates the potential application of LLMs to slot filling with noisy ASR transcriptions, via both in-context learning and task-specific fine-tuning. Dedicated prompt designs and fine-tuning approaches are proposed to improve the robustness of LLMs for slot filling with noisy ASR transcriptions. Moreover, a linearised knowledge injection (LKI) scheme is also proposed to integrate dynamic external knowledge into LLMs. Experiments were performed on SLURP to quantify the performance of LLMs, including GPT-3.5-turbo, GPT-4, LLaMA-13B and Vicuna-13B (v1.1 and v1.5) with different ASR error rates. The use of the proposed fine-tuning together with the LKI scheme for LLaMA-13B achieved an 8.3% absolute SLU-F1 improvement compared to the strong Flan-T5-base baseline system on a limited data setup.

Shutong Feng, Guangzhi Sun, Nurul Lubis et al.

Affect recognition, encompassing emotions, moods, and feelings, plays a pivotal role in human communication. In the realm of conversational artificial intelligence, the ability to discern and respond to human affective cues is a critical factor for creating engaging and empathetic interactions. This study investigates the capacity of large language models (LLMs) to recognise human affect in conversations, with a focus on both open-domain chit-chat dialogues and task-oriented dialogues. Leveraging three diverse datasets, namely IEMOCAP (Busso et al., 2008), EmoWOZ (Feng et al., 2022), and DAIC-WOZ (Gratch et al., 2014), covering a spectrum of dialogues from casual conversations to clinical interviews, we evaluate and compare LLMs' performance in affect recognition. Our investigation explores the zero-shot and few-shot capabilities of LLMs through in-context learning as well as their model capacities through task-specific fine-tuning. Additionally, this study takes into account the potential impact of automatic speech recognition errors on LLM predictions. With this work, we aim to shed light on the extent to which LLMs can replicate human-like affect recognition capabilities in conversations.

Adel Moumen, Guangzhi Sun, Philip C. Woodland

Spoken Language Models (SLMs) aim to learn linguistic competence directly from speech using discrete units, widening access to Natural Language Processing (NLP) technologies for languages with limited written resources. However, progress has been largely English-centric due to scarce spoken evaluation benchmarks and training data, making cross-lingual learning difficult. We present a cross-lingual interleaving method that mixes speech tokens across languages without textual supervision. We also release an EN-FR training dataset, TinyStories (~42k hours), together with EN-FR spoken StoryCloze and TopicCloze benchmarks for cross-lingual semantic evaluation, both synthetically generated using GPT-4. On 360M and 1B SLMs under matched training-token budgets, interleaving improves monolingual semantic accuracy, enables robust cross-lingual continuation, and strengthens cross-lingual hidden-state alignment. Taken together, these results indicate that cross-lingual interleaving is a simple, scalable route to building multilingual SLMs that understand and converse across languages. All resources will be made open-source to support reproducibility.

Theodor Nguyen, Guangzhi Sun, Xianrui Zheng et al.

We propose DiffSpEx, a generative target speaker extraction method based on score-based generative modelling through stochastic differential equations. DiffSpEx deploys a continuous-time stochastic diffusion process in the complex short-time Fourier transform domain, starting from the target speaker source and converging to a Gaussian distribution centred on the mixture of sources. For the reverse-time process, a parametrised score function is conditioned on a target speaker embedding to extract the target speaker from the mixture of sources. We utilise ECAPA-TDNN target speaker embeddings and condition the score function alternately on the SDE time embedding and the target speaker embedding. The potential of DiffSpEx is demonstrated with the WSJ0-2mix dataset, achieving an SI-SDR of 12.9 dB and a NISQA score of 3.56. Moreover, we show that fine-tuning a pre-trained DiffSpEx model to a specific speaker further improves performance, enabling personalisation in target speaker extraction.

Qiuming Zhao, Guangzhi Sun, Chao Zhang et al.

Recent end-to-end automatic speech recognition (ASR) models have become increasingly larger, making them particularly challenging to be deployed on resource-constrained devices. Model quantisation is an effective solution that sometimes causes the word error rate (WER) to increase. In this paper, a novel strategy of personalisation for a quantised model (PQM) is proposed, which combines speaker adaptive training (SAT) with model quantisation to improve the performance of heavily compressed models. Specifically, PQM uses a 4-bit NormalFloat Quantisation (NF4) approach for model quantisation and low-rank adaptation (LoRA) for SAT. Experiments have been performed on the LibriSpeech and the TED-LIUM 3 corpora. Remarkably, with a 7x reduction in model size and 1% additional speaker-specific parameters, 15.1% and 23.3% relative WER reductions were achieved on quantised Whisper and Conformer-based attention-based encoder-decoder ASR models respectively, comparing to the original full precision models.

Nineli Lashkarashvili, Wen Wu, Guangzhi Sun et al.

Foundation models have shown superior performance for speech emotion recognition (SER). However, given the limited data in emotion corpora, finetuning all parameters of large pre-trained models for SER can be both resource-intensive and susceptible to overfitting. This paper investigates parameter-efficient finetuning (PEFT) for SER. Various PEFT adaptors are systematically studied for both classification of discrete emotion categories and prediction of dimensional emotional attributes. The results demonstrate that the combination of PEFT methods surpasses full finetuning with a significant reduction in the number of trainable parameters. Furthermore, a two-stage adaptation strategy is proposed to adapt models trained on acted emotion data, which is more readily available, to make the model more adept at capturing natural emotional expressions. Both intra- and cross-corpus experiments validate the efficacy of the proposed approach in enhancing the performance on both the source and target domains.

Xiaoliang Luo, Guangzhi Sun, Bradley C. Love

Recently, large language models (LLMs) have outperformed human experts in predicting the results of neuroscience experiments (Luo et al., 2024). What is the basis for this performance? One possibility is that statistical patterns in that specific scientific literature, as opposed to emergent reasoning abilities arising from broader training, underlie LLMs' performance. To evaluate this possibility, we trained (next word prediction) a relatively small 124M-parameter GPT-2 model on 1.3 billion tokens of domain-specific knowledge. Despite being orders of magnitude smaller than larger LLMs trained on trillions of tokens, small models achieved expert-level performance in predicting neuroscience results. Small models trained on the neuroscience literature succeeded when they were trained from scratch using a tokenizer specifically trained on neuroscience text or when the neuroscience literature was used to finetune a pretrained GPT-2. Our results indicate that expert-level performance may be attained by even small LLMs through domain-specific, auto-regressive training approaches.

Guangzhi Sun, Chao Zhang, Philip C. Woodland

Contextual knowledge is important for real-world automatic speech recognition (ASR) applications. In this paper, a novel tree-constrained pointer generator (TCPGen) component is proposed that incorporates such knowledge as a list of biasing words into both attention-based encoder-decoder and transducer end-to-end ASR models in a neural-symbolic way. TCPGen structures the biasing words into an efficient prefix tree to serve as its symbolic input and creates a neural shortcut between the tree and the final ASR output distribution to facilitate recognising biasing words during decoding. Systems were trained and evaluated on the Librispeech corpus where biasing words were extracted at the scales of an utterance, a chapter, or a book to simulate different application scenarios. Experimental results showed that TCPGen consistently improved word error rates (WERs) compared to the baselines, and in particular, achieved significant WER reductions on the biasing words. TCPGen is highly efficient: it can handle 5,000 biasing words and distractors and only add a small overhead to memory use and computation cost.

G. Sun, C. Zhang, P. C. Woodland

The effective incorporation of cross-utterance information has the potential to improve language models (LMs) for automatic speech recognition (ASR). To extract more powerful and robust cross-utterance representations for the Transformer LM (TLM), this paper proposes the R-TLM which uses hidden states in a long short-term memory (LSTM) LM. To encode the cross-utterance information, the R-TLM incorporates an LSTM module together with a segment-wise recurrence in some of the Transformer blocks. In addition to the LSTM module output, a shortcut connection using a fusion layer that bypasses the LSTM module is also investigated. The proposed system was evaluated on the AMI meeting corpus, the Eval2000 and the RT03 telephone conversation evaluation sets. The best R-TLM achieved 0.9%, 0.6%, and 0.8% absolute WER reductions over the single-utterance TLM baseline, and 0.5%, 0.3%, 0.2% absolute WER reductions over a strong cross-utterance TLM baseline on the AMI evaluation set, Eval2000 and RT03 respectively. Improvements on Eval2000 and RT03 were further supported by significance tests. R-TLMs were found to have better LM scores on words where recognition errors are more likely to occur. The R-TLM WER can be further reduced by interpolation with an LSTM-LM.

G. Sun, D. Liu, C. Zhang et al.

Recent speaker diarisation systems often convert variable length speech segments into fixed-length vector representations for speaker clustering, which are known as speaker embeddings. In this paper, the content-aware speaker embeddings (CASE) approach is proposed, which extends the input of the speaker classifier to include not only acoustic features but also their corresponding speech content, via phone, character, and word embeddings. Compared to alternative methods that leverage similar information, such as multitask or adversarial training, CASE factorises automatic speech recognition (ASR) from speaker recognition to focus on modelling speaker characteristics and correlations with the corresponding content units to derive more expressive representations. CASE is evaluated for speaker re-clustering with a realistic speaker diarisation setup using the AMI meeting transcription dataset, where the content information is obtained by performing ASR based on an automatic segmentation. Experimental results showed that CASE achieved a 17.8% relative speaker error rate reduction over conventional methods.

Guangzhi Sun, Chao Zhang, Phil Woodland

Significant progress has recently been made in speaker diarisation after the introduction of d-vectors as speaker embeddings extracted from neural network (NN) speaker classifiers for clustering speech segments. To extract better-performing and more robust speaker embeddings, this paper proposes a c-vector method by combining multiple sets of complementary d-vectors derived from systems with different NN components. Three structures are used to implement the c-vectors, namely 2D self-attentive, gated additive, and bilinear pooling structures, relying on attention mechanisms, a gating mechanism, and a low-rank bilinear pooling mechanism respectively. Furthermore, a neural-based single-pass speaker diarisation pipeline is also proposed in this paper, which uses NNs to achieve voice activity detection, speaker change point detection, and speaker embedding extraction. Experiments and detailed analyses are conducted on the challenging AMI and NIST RT05 datasets which consist of real meetings with 4--10 speakers and a wide range of acoustic conditions. For systems trained on the AMI training set, relative speaker error rate (SER) reductions of 13% and 29% are obtained by using c-vectors instead of d-vectors on the AMI dev and eval sets respectively, and a relative reduction of 15% in SER is observed on RT05, which shows the robustness of the proposed methods. By incorporating VoxCeleb data into the training set, the best c-vector system achieved 7%, 17% and16% relative SER reduction compared to the d-vector on the AMI dev, eval, and RT05 sets respectively

G. Sun, C. Zhang, P. C. Woodland

The effective exploitation of richer contextual information in language models (LMs) is a long-standing research problem for automatic speech recognition (ASR). A cross-utterance LM (CULM) is proposed in this paper, which augments the input to a standard long short-term memory (LSTM) LM with a context vector derived from past and future utterances using an extraction network. The extraction network uses another LSTM to encode surrounding utterances into vectors which are integrated into a context vector using either a projection of LSTM final hidden states, or a multi-head self-attentive layer. In addition, an acoustic error sampling technique is proposed to reduce the mismatch between training and test-time. This is achieved by considering possible ASR errors into the model training procedure, and can therefore improve the word error rate (WER). Experiments performed on both AMI and Switchboard datasets show that CULMs outperform the LSTM LM baseline WER. In particular, the CULM with a self-attentive layer-based extraction network and acoustic error sampling achieves 0.6% absolute WER reduction on AMI, 0.3% WER reduction on the Switchboard part and 0.9% WER reduction on the Callhome part of Eval2000 test set over the respective baselines.

Guangzhi Sun, Yu Zhang, Ron J. Weiss et al.

This paper proposes a hierarchical, fine-grained and interpretable latent variable model for prosody based on the Tacotron 2 text-to-speech model. It achieves multi-resolution modeling of prosody by conditioning finer level representations on coarser level ones. Additionally, it imposes hierarchical conditioning across all latent dimensions using a conditional variational auto-encoder (VAE) with an auto-regressive structure. Evaluation of reconstruction performance illustrates that the new structure does not degrade the model while allowing better interpretability. Interpretations of prosody attributes are provided together with the comparison between word-level and phone-level prosody representations. Moreover, both qualitative and quantitative evaluations are used to demonstrate the improvement in the disentanglement of the latent dimensions.

Guangzhi Sun, Yu Zhang, Ron J. Weiss et al.

Recent neural text-to-speech (TTS) models with fine-grained latent features enable precise control of the prosody of synthesized speech. Such models typically incorporate a fine-grained variational autoencoder (VAE) structure, extracting latent features at each input token (e.g., phonemes). However, generating samples with the standard VAE prior often results in unnatural and discontinuous speech, with dramatic prosodic variation between tokens. This paper proposes a sequential prior in a discrete latent space which can generate more naturally sounding samples. This is accomplished by discretizing the latent features using vector quantization (VQ), and separately training an autoregressive (AR) prior model over the result. We evaluate the approach using listening tests, objective metrics of automatic speech recognition (ASR) performance, and measurements of prosody attributes. Experimental results show that the proposed model significantly improves the naturalness in random sample generation. Furthermore, initial experiments demonstrate that randomly sampling from the proposed model can be used as data augmentation to improve the ASR performance.

Guangzhi Sun, Chao Zhang, Phil Woodland

Speaker diarisation systems often cluster audio segments using speaker embeddings such as i-vectors and d-vectors. Since different types of embeddings are often complementary, this paper proposes a generic framework to improve performance by combining them into a single embedding, referred to as a c-vector. This combination uses a 2-dimensional (2D) self-attentive structure, which extends the standard self-attentive layer by averaging not only across time but also across different types of embeddings. Two types of 2D self-attentive structure in this paper are the simultaneous combination and the consecutive combination, adopting a single and multiple self-attentive layers respectively. The penalty term in the original self-attentive layer which is jointly minimised with the objective function to encourage diversity of annotation vectors is also modified to obtain not only different local peaks but also the overall trends in the multiple annotation vectors. Experiments on the AMI meeting corpus show that our modified penalty term improves the d- vector relative speaker error rate (SER) by 6% and 21% for d-vector systems, and a 10% further relative SER reduction can be obtained using the c-vector from our best 2D self-attentive structure.