分享
Existing research suggests that automatic speech recognition (ASR) models can benefit from additional contexts (e.g., contact lists, user specified vocabulary). Rare words and named entities can be better recognized with contexts. In this work, we propose two simple yet effective techniques to improve context-aware ASR models. First, we inject contexts into the encoders at an early stage instead of merely at their last layers. Second, to enforce the model to leverage the contexts during training, we perturb the reference transcription with alternative spellings so that the model learns to rely on the contexts to make correct predictions. On LibriSpeech, our techniques together reduce the rare word error rate by 60% and 25% relatively compared to no biasing and shallow fusion, making the new state-of-the-art performance. On SPGISpeech and a real-world dataset ConEC, our techniques also yield good improvements over the baselines.
Knowledge distillation plays a key role in compressing the Large Language Models (LLMs), which boosts a small-size student model under large teacher models' guidance. However, existing LLM distillation methods overly rely on student-generated outputs, which may introduce generation errors and misguide the distillation process. Moreover, the distillation loss functions introduced in previous art struggle to align the most informative part due to the complex distribution of LLMs' outputs. To address these problems, we propose a multi-granularity semantic revision method for LLM distillation. At the sequence level, we propose a sequence correction and re-generation (SCRG) strategy. SCRG first calculates the semantic cognitive difference between the teacher and student to detect the error token, then corrects it with the teacher-generated one, and re-generates the sequence to reduce generation errors and enhance generation diversity. At the token level, we design a distribution adaptive clipping Kullback-Leibler (DAC-KL) loss as the distillation objective function. DAC-KL loss exploits a learnable sub-network to adaptively extract semantically dense areas from the teacher's output, avoiding the interference of redundant information in the distillation process. Finally, at the span level, we leverage the span priors of a sequence to compute the probability correlations within spans, and constrain the teacher and student's probability correlations to be consistent, further enhancing the transfer of semantic information. Extensive experiments across different model families with parameters ranging from 0.1B to 13B demonstrate the superiority of our method compared to existing methods.
Words often carry different meanings for people from diverse backgrounds. Today's era of social polarization demands that we choose words carefully to prevent miscommunication, especially in political communication and journalism. To address this issue, we introduce the Bridging Dictionary, an interactive tool designed to illuminate how words are perceived by people with different political views. The Bridging Dictionary includes a static, printable document featuring 796 terms with summaries generated by a large language model. These summaries highlight how the terms are used distinctively by Republicans and Democrats. Additionally, the Bridging Dictionary offers an interactive interface that lets users explore selected words, visualizing their frequency, sentiment, summaries, and examples across political divides. We present a use case for journalists and emphasize the importance of human agency and trust in further enhancing this tool. The deployed version of Bridging Dictionary is available at https://dictionary.ccc-mit.org/.
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.
When it comes to authentication in speaker verification systems, not all utterances are created equal. It is essential to estimate the quality of test utterances in order to account for varying acoustic conditions. In addition to the net-speech duration of an utterance, it is observed in this paper that phonetic richness is also a key indicator of utterance quality, playing a significant role in accurate speaker verification. Several phonetic histogram based formulations of phonetic richness are explored using transcripts obtained from an automatic speaker recognition system. The proposed phonetic richness measure is found to be positively correlated with voice authentication scores across evaluation benchmarks. Additionally, the proposed measure in combination with net speech helps in calibrating the speaker verification scores, obtaining a relative EER improvement of 5.8% on the Voxceleb1 evaluation protocol. The proposed phonetic richness based calibration provides higher benefit for short utterances with repeated words.
Large language models (LLMs) are quickly being adopted in a wide range of learning experiences, especially via ubiquitous and broadly accessible chat interfaces like ChatGPT and Copilot. This type of interface is readily available to students and teachers around the world, yet relatively little research has been done to assess the impact of such generic tools on student learning. Coding education is an interesting test case, both because LLMs have strong performance on coding tasks, and because LLM-powered support tools are rapidly becoming part of the workflow of professional software engineers. To help understand the impact of generic LLM use on coding education, we conducted a large-scale randomized control trial with 5,831 students from 146 countries in an online coding class in which we provided some students with access to a chat interface with GPT-4. We estimate positive benefits on exam performance for adopters, the students who used the tool, but over all students, the advertisement of GPT-4 led to a significant average decrease in exam participation. We observe similar decreases in other forms of course engagement. However, this decrease is modulated by the student's country of origin. Offering access to LLMs to students from low human development index countries increased their exam participation rate on average. Our results suggest there may be promising benefits to using LLMs in an introductory coding class, but also potential harms for engagement, which makes their longer term impact on student success unclear. Our work highlights the need for additional investigations to help understand the potential impact of future adoption and integration of LLMs into classrooms.
This dataset includes 6823 thermal images captured using a UNI-T UTi165A camera for face detection, recognition, and emotion analysis. It consists of 2485 facial recognition images depicting emotions (happy, sad, angry, natural, surprised), 2054 images for face recognition, and 2284 images for face detection. The dataset covers various conditions, color palettes, shooting angles, and zoom levels, with a temperature range of -10{\deg}C to 400{\deg}C and a resolution of 19,200 pixels. It serves as a valuable resource for advancing thermal imaging technology, aiding in algorithm development, and benchmarking for facial recognition across different palettes. Additionally, it contributes to facial motion recognition, fostering interdisciplinary collaboration in computer vision, psychology, and neuroscience. The dataset promotes transparency in thermal face detection and recognition research, with applications in security, healthcare, and human-computer interaction.
The proliferation of e-commerce and urbanization has significantly intensified delivery operations in urban areas, boosting the volume and complexity of delivery demand. Data-driven predictive methods, especially those utilizing machine learning techniques, have emerged to handle these complexities in urban delivery demand management problems. One particularly pressing problem that has not yet been sufficiently studied is the joint estimation and prediction of city-wide delivery demand. To this end, we formulate this problem as a graph-based spatiotemporal learning task. First, a message-passing neural network model is formalized to capture the interaction between demand patterns of associated regions. Second, by exploiting recent advances in large language models, we extract general geospatial knowledge encodings from the unstructured locational data and integrate them into the demand predictor. Last, to encourage the cross-city transferability of the model, an inductive training scheme is developed in an end-to-end routine. Extensive empirical results on two real-world delivery datasets, including eight cities in China and the US, demonstrate that our model significantly outperforms state-of-the-art baselines in these challenging tasks.
Electroencephalography (EEG) provides reliable indications of human cognition and mental states. Accurate emotion recognition from EEG remains challenging due to signal variations among individuals and across measurement sessions. To address these challenges, we introduce a multi-source dynamic contrastive domain adaptation method (MS-DCDA), which models coarse-grained inter-domain and fine-grained intra-class adaptations through a multi-branch contrastive neural network and contrastive sub-domain discrepancy learning. Our model leverages domain knowledge from each individual source and a complementary source ensemble and uses dynamically weighted learning to achieve an optimal tradeoff between domain transferability and discriminability. The proposed MS-DCDA model was evaluated using the SEED and SEED-IV datasets, achieving respectively the highest mean accuracies of $90.84\%$ and $78.49\%$ in cross-subject experiments as well as $95.82\%$ and $82.25\%$ in cross-session experiments. Our model outperforms several alternative domain adaptation methods in recognition accuracy, inter-class margin, and intra-class compactness. Our study also suggests greater emotional sensitivity in the frontal and parietal brain lobes, providing insights for mental health interventions, personalized medicine, and development of preventive strategies.
Video deblurring is a challenging task that aims to recover sharp sequences from blur and noisy observations. The image-formation model plays a crucial role in traditional model-based methods, constraining the possible solutions. However, this is only the case for some deep learning-based methods. Despite deep-learning models achieving better results, traditional model-based methods remain widely popular due to their flexibility. An increasing number of scholars combine the two to achieve better deblurring performance. This paper proposes introducing knowledge of the image-formation model into a deep learning network by using the pseudo-inverse of the blur. We use a deep network to fit the blurring and estimate pseudo-inverse. Then, we use this estimation, combined with a variational deep-learning network, to deblur the video sequence. Notably, our experimental results demonstrate that such modifications can significantly improve the performance of deep learning models for video deblurring. Furthermore, our experiments on different datasets achieved notable performance improvements, proving that our proposed method can generalize to different scenarios and cameras.
Target speaker extraction aims to isolate the voice of a specific speaker from mixed speech. Traditionally, this process has relied on extracting a speaker embedding from a reference speech, necessitating a speaker recognition model. However, identifying an appropriate speaker recognition model can be challenging, and using the target speaker embedding as reference information may not be optimal for target speaker extraction tasks. This paper introduces a Universal Speaker Embedding-Free Target Speaker Extraction (USEF-TSE) framework that operates without relying on speaker embeddings. USEF-TSE utilizes a multi-head cross-attention mechanism as a frame-level target speaker feature extractor. This innovative approach allows mainstream speaker extraction solutions to bypass the dependency on speaker recognition models and to fully leverage the information available in the enrollment speech, including speaker characteristics and contextual details. Additionally, USEF-TSE can seamlessly integrate with any time-domain or time-frequency domain speech separation model to achieve effective speaker extraction. Experimental results show that our proposed method achieves state-of-the-art (SOTA) performance in terms of Scale-Invariant Signal-to-Distortion Ratio (SI-SDR) on the WSJ0-2mix, WHAM!, and WHAMR! datasets, which are standard benchmarks for monaural anechoic, noisy and noisy-reverberant two-speaker speech separation and speaker extraction.
In recent years, there has been significant progress in Text-to-Speech (TTS) synthesis technology, enabling the high-quality synthesis of voices in common scenarios. In unseen situations, adaptive TTS requires a strong generalization capability to speaker style characteristics. However, the existing adaptive methods can only extract and integrate coarse-grained timbre or mixed rhythm attributes separately. In this paper, we propose AS-Speech, an adaptive style methodology that integrates the speaker timbre characteristics and rhythmic attributes into a unified framework for text-to-speech synthesis. Specifically, AS-Speech can accurately simulate style characteristics through fine-grained text-based timbre features and global rhythm information, and achieve high-fidelity speech synthesis through the diffusion model. Experiments show that the proposed model produces voices with higher naturalness and similarity in terms of timbre and rhythm compared to a series of adaptive TTS models.