Incremental text-to-speech, also known as streaming TTS, has been increasingly applied to online speech applications that require ultra-low response latency to provide an optimal user experience. However, most of the existing speech synthesis pipelines deployed on GPU are still non-incremental, which uncovers limitations in high-concurrency scenarios, especially when the pipeline is built with end-to-end neural network models. To address this issue, we present a highly efficient approach to perform real-time incremental TTS on GPUs with Instant Request Pooling and Module-wise Dynamic Batching. Experimental results demonstrate that the proposed method is capable of producing high-quality speech with a first-chunk latency lower than 80ms under 100 QPS on a single NVIDIA A10 GPU and significantly outperforms the non-incremental twin in both concurrency and latency. Our work reveals the effectiveness of high-performance incremental TTS on GPUs.
We enhance the vanilla adversarial training method for unsupervised Automatic Speech Recognition (ASR) by a diffusion-GAN. Our model (1) injects instance noises of various intensities to the generator's output and unlabeled reference text which are sampled from pretrained phoneme language models with a length constraint, (2) asks diffusion timestep-dependent discriminators to separate them, and (3) back-propagates the gradients to update the generator. Word/phoneme error rate comparisons with wav2vec-U under Librispeech (3.1% for test-clean and 5.6% for test-other), TIMIT and MLS datasets, show that our enhancement strategies work effectively.
We propose a method for editing NeRF scenes with text-instructions. Given a NeRF of a scene and the collection of images used to reconstruct it, our method uses an image-conditioned diffusion model (InstructPix2Pix) to iteratively edit the input images while optimizing the underlying scene, resulting in an optimized 3D scene that respects the edit instruction. We demonstrate that our proposed method is able to edit large-scale, real-world scenes, and is able to accomplish more realistic, targeted edits than prior work.
Traditional semantic image search methods aim to retrieve images that match the meaning of the text query. However, these methods typically search for objects on the whole image, without considering the localization of objects within the image. This paper presents an extension of existing object detection models for semantic image search that considers the semantic alignment between object proposals and text queries, with a focus on searching for objects within images. The proposed model uses a single feature extractor, a pre-trained Convolutional Neural Network, and a transformer encoder to encode the text query. Proposal-text alignment is performed using contrastive learning, producing a score for each proposal that reflects its semantic alignment with the text query. The Region Proposal Network (RPN) is used to generate object proposals, and the end-to-end training process allows for an efficient and effective solution for semantic image search. The proposed model was trained end-to-end, providing a promising solution for semantic image search that retrieves images that match the meaning of the text query and generates semantically relevant object proposals.
To leverage machine learning in any decision-making process, one must convert the given knowledge (for example, natural language, unstructured text) into representation vectors that can be understood and processed by machine learning model in their compatible language and data format. The frequently encountered difficulty is, however, the given knowledge is not rich or reliable enough in the first place. In such cases, one seeks to fuse side information from a separate domain to mitigate the gap between good representation learning and the scarce knowledge in the domain of interest. This approach is named Cross-Domain Knowledge Transfer. It is crucial to study the problem because of the commonality of scarce knowledge in many scenarios, from online healthcare platform analyses to financial market risk quantification, leaving an obstacle in front of us benefiting from automated decision making. From the machine learning perspective, the paradigm of semi-supervised learning takes advantage of large amount of data without ground truth and achieves impressive learning performance improvement. It is adopted in this dissertation for cross-domain knowledge transfer. (to be continued)
Advances in natural language generation (NLG) have resulted in machine generated text that is increasingly difficult to distinguish from human authored text. Powerful open-source models are freely available, and user-friendly tools democratizing access to generative models are proliferating. The great potential of state-of-the-art NLG systems is tempered by the multitude of avenues for abuse. Detection of machine generated text is a key countermeasure for reducing abuse of NLG models, with significant technical challenges and numerous open problems. We provide a survey that includes both 1) an extensive analysis of threat models posed by contemporary NLG systems, and 2) the most complete review of machine generated text detection methods to date. This survey places machine generated text within its cybersecurity and social context, and provides strong guidance for future work addressing the most critical threat models, and ensuring detection systems themselves demonstrate trustworthiness through fairness, robustness, and accountability.
Reinforcement learning (RL) has been widely used in text generation to alleviate the exposure bias issue or to utilize non-parallel datasets. The reward function plays an important role in making RL training successful. However, previous reward functions are typically task-specific and sparse, restricting the use of RL. In our work, we propose a task-agnostic approach that derives a step-wise reward function directly from a model trained with teacher forcing. We additionally propose a simple modification to stabilize the RL training on non-parallel datasets with our induced reward function. Empirical results show that our method outperforms self-training and reward regression methods on several text generation tasks, confirming the effectiveness of our reward function.
BERT is a widely used pre-trained model in natural language processing. However, since BERT is quadratic to the text length, the BERT model is difficult to be used directly on the long-text corpus. In some fields, the collected text data may be quite long, such as in the health care field. Therefore, to apply the pre-trained language knowledge of BERT to long text, in this paper, imitating the skimming-intensive reading method used by humans when reading a long paragraph, the Skimming-Intensive Model (SkIn) is proposed. It can dynamically select the critical information in the text so that the sentence input into the BERT-Base model is significantly shortened, which can effectively save the cost of the classification algorithm. Experiments show that the SkIn method has achieved superior accuracy than the baselines on long-text classification datasets in the medical field, while its time and space requirements increase linearly with the text length, alleviating the time and space overflow problem of basic BERT on long-text data.
Denoising diffusion models are a recent class of generative models which achieve state-of-the-art results in many domains such as unconditional image generation and text-to-speech tasks. They consist of a noising process destroying the data and a backward stage defined as the time-reversal of the noising diffusion. Building on their success, diffusion models have recently been extended to the Riemannian manifold setting. Yet, these Riemannian diffusion models require geodesics to be defined for all times. While this setting encompasses many important applications, it does not include manifolds defined via a set of inequality constraints, which are ubiquitous in many scientific domains such as robotics and protein design. In this work, we introduce two methods to bridge this gap. First, we design a noising process based on the logarithmic barrier metric induced by the inequality constraints. Second, we introduce a noising process based on the reflected Brownian motion. As existing diffusion model techniques cannot be applied in this setting, we derive new tools to define such models in our framework. We empirically demonstrate the applicability of our methods to a number of synthetic and real-world tasks, including the constrained conformational modelling of protein backbones and robotic arms.
The core challenge in numerous real-world applications is to match an inquiry to the best document from a mutable and finite set of candidates. Existing industry solutions, especially latency-constrained services, often rely on similarity algorithms that sacrifice quality for speed. In this paper we introduce a generic semantic learning-to-rank framework, Self-training Semantic Cross-attention Ranking (sRank). This transformer-based framework uses linear pairwise loss with mutable training batch sizes and achieves quality gains and high efficiency, and has been applied effectively to show gains on two industry tasks at Microsoft over real-world large-scale data sets: Smart Reply (SR) and Ambient Clinical Intelligence (ACI). In Smart Reply, $sRank$ assists live customers with technical support by selecting the best reply from predefined solutions based on consumer and support agent messages. It achieves 11.7% gain in offline top-one accuracy on the SR task over the previous system, and has enabled 38.7% time reduction in composing messages in telemetry recorded since its general release in January 2021. In the ACI task, sRank selects relevant historical physician templates that serve as guidance for a text summarization model to generate higher quality medical notes. It achieves 35.5% top-one accuracy gain, along with 46% relative ROUGE-L gain in generated medical notes.