The cross-speaker emotion transfer task in TTS particularly aims to synthesize speech for a target speaker with the emotion transferred from reference speech recorded by another (source) speaker. During the emotion transfer process, the identity info
rmation of the source speaker could also affect the synthesized results, resulting in the issue of speaker leakage. This paper proposes a new method with the aim to synthesize controllable emotional expressive speech and meanwhile maintain the target speakers identity in the cross-speaker emotion TTS task. The proposed method is a Tacotron2-based framework with the emotion embedding as the conditioning variable to provide emotion information. Two emotion disentangling modules are contained in our method to 1) get speaker-independent and emotion-discriminative embedding, and 2) explicitly constrain the emotion and speaker identity of synthetic speech to be that as expected. Moreover, we present an intuitive method to control the emotional strength in the synthetic speech for the target speaker. Specifically, the learned emotion embedding is adjusted with a flexible scalar value, which allows controlling the emotion strength conveyed by the embedding. Extensive experiments have been conducted on a Mandarin disjoint corpus, and the results demonstrate that the proposed method is able to synthesize reasonable emotional speech for the target speaker. Compared to the state-of-the-art reference embedding learned methods, our method gets the best performance on the cross-speaker emotion transfer task, indicating that our method achieves the new state-of-the-art performance on learning the speaker-independent emotion embedding. Furthermore, the strength ranking test and pitch trajectories plots demonstrate that the proposed method can effectively control the emotion strength, leading to prosody-diverse synthetic speech.
With the rapid increase of multimedia data, a large body of literature has emerged to work on multimodal summarization, the majority of which target at refining salient information from textual and visual modalities to output a pictorial summary with
the most relevant images. Existing methods mostly focus on either extractive or abstractive summarization and rely on qualified image captions to build image references. We are the first to propose a Unified framework for Multimodal Summarization grounding on BART, UniMS, that integrates extractive and abstractive objectives, as well as selecting the image output. Specially, we adopt knowledge distillation from a vision-language pretrained model to improve image selection, which avoids any requirement on the existence and quality of image captions. Besides, we introduce a visual guided decoder to better integrate textual and visual modalities in guiding abstractive text generation. Results show that our best model achieves a new state-of-the-art result on a large-scale benchmark dataset. The newly involved extractive objective as well as the knowledge distillation technique are proven to bring a noticeable improvement to the multimodal summarization task.
As labeling cost for different modules in task-oriented dialog (ToD) systems is high, a major challenge in practice is to learn different tasks with the least amount of labeled data. Recently, prompting methods over pre-trained language models (PLMs)
have shown promising results for few-shot learning in ToD. To better utilize the power of PLMs, this paper proposes Comprehensive Instruction (CINS) that exploits PLMs with extra task-specific instructions. We design a schema (definition, constraint, prompt) of instructions and their customized realizations for three important downstream tasks in ToD, i.e. intent classification, dialog state tracking, and natural language generation. A sequence-to-sequence model (T5) is adopted to solve these three tasks in a unified framework. Extensive experiments are conducted on these ToD tasks in realistic few-shot learning scenarios with small validation data. Empirical results demonstrate that the proposed CINS approach consistently improves techniques that finetune PLMs with raw input or short prompts.
In this paper, we present explicit and computable error bounds for the asymptotic expansions of Hermite polynomials with Plancherel-Rotach scale. Three cases, depending on whether the scaled variable lies in the outer or oscillatory interval, or it i
s the turning point, are considered respectively. We introduce the branch cut technique to express the error term as an integral on the contour taking as the one-sided limit of curves approaching the branch cut. This new technique enables us to derive simple formulas for the error bounds in terms of elementary functions.
We present a reinforcement learning (RL) approach for robust optimisation of risk-aware performance criteria. To allow agents to express a wide variety of risk-reward profiles, we assess the value of a policy using rank dependent expected utility (RD
EU). RDEU allows the agent to seek gains, while simultaneously protecting themselves against downside events. To robustify optimal policies against model uncertainty, we assess a policy not by its distribution, but rather, by the worst possible distribution that lies within a Wasserstein ball around it. Thus, our problem formulation may be viewed as an actor choosing a policy (the outer problem), and the adversary then acting to worsen the performance of that strategy (the inner problem). We develop explicit policy gradient formulae for the inner and outer problems, and show its efficacy on three prototypical financial problems: robust portfolio allocation, optimising a benchmark, and statistical arbitrage
The paper proposes a Dynamic ResBlock Generative Adversarial Network (DRB-GAN) for artistic style transfer. The style code is modeled as the shared parameters for Dynamic ResBlocks connecting both the style encoding network and the style transfer net
work. In the style encoding network, a style class-aware attention mechanism is used to attend the style feature representation for generating the style codes. In the style transfer network, multiple Dynamic ResBlocks are designed to integrate the style code and the extracted CNN semantic feature and then feed into the spatial window Layer-Instance Normalization (SW-LIN) decoder, which enables high-quality synthetic images with artistic style transfer. Moreover, the style collection conditional discriminator is designed to equip our DRB-GAN model with abilities for both arbitrary style transfer and collection style transfer during the training stage. No matter for arbitrary style transfer or collection style transfer, extensive experiments strongly demonstrate that our proposed DRB-GAN outperforms state-of-the-art methods and exhibits its superior performance in terms of visual quality and efficiency. Our source code is available at color{magenta}{url{https://github.com/xuwenju123/DRB-GAN}}.
Code representation learning, which aims to encode the semantics of source code into distributed vectors, plays an important role in recent deep-learning-based models for code intelligence. Recently, many pre-trained language models for source code (
e.g., CuBERT and CodeBERT) have been proposed to model the context of code and serve as a basis for downstream code intelligence tasks such as code search, code clone detection, and program translation. Current approaches typically consider the source code as a plain sequence of tokens, or inject the structure information (e.g., AST and data-flow) into the sequential model pre-training. To further explore the properties of programming languages, this paper proposes SynCoBERT, a syntax-guided multi-modal contrastive pre-training approach for better code representations. Specially, we design two novel pre-training objectives originating from the symbolic and syntactic properties of source code, i.e., Identifier Prediction (IP) and AST Edge Prediction (TEP), which are designed to predict identifiers, and edges between two nodes of AST, respectively. Meanwhile, to exploit the complementary information in semantically equivalent modalities (i.e., code, comment, AST) of the code, we propose a multi-modal contrastive learning strategy to maximize the mutual information among different modalities. Extensive experiments on four downstream tasks related to code intelligence show that SynCoBERT advances the state-of-the-art with the same pre-training corpus and model size.
Automatically generating videos in which synthesized speech is synchronized with lip movements in a talking head has great potential in many human-computer interaction scenarios. In this paper, we present an automatic method to generate synchronized
speech and talking-head videos on the basis of text and a single face image of an arbitrary person as input. In contrast to previous text-driven talking head generation methods, which can only synthesize the voice of a specific person, the proposed method is capable of synthesizing speech for any person that is inaccessible in the training stage. Specifically, the proposed method decomposes the generation of synchronized speech and talking head videos into two stages, i.e., a text-to-speech (TTS) stage and a speech-driven talking head generation stage. The proposed TTS module is a face-conditioned multi-speaker TTS model that gets the speaker identity information from face images instead of speech, which allows us to synthesize a personalized voice on the basis of the input face image. To generate the talking head videos from the face images, a facial landmark-based method that can predict both lip movements and head rotations is proposed. Extensive experiments demonstrate that the proposed method is able to generate synchronized speech and talking head videos for arbitrary persons and non-persons. Synthesized speech shows consistency with the given face regarding to the synthesized voices timbre and ones appearance in the image, and the proposed landmark-based talking head method outperforms the state-of-the-art landmark-based method on generating natural talking head videos.
In this paper, we propose a direct parallel-in-time (PinT) algorithm for time-dependent problems with first- or second-order derivative. We use a second-order boundary value method as the time integrator that leads to a tridiagonal time discretizatio
n matrix. Instead of solving the corresponding all-at-once system iteratively, we diagonalize the time discretization matrix, which yields a direct parallel implementation across all time levels. A crucial issue on this methodology is how the condition number of the eigenvector matrix $V$ grows as $n$ is increased, where $n$ is the number of time levels. A large condition number leads to large roundoff error in the diagonalization procedure, which could seriously pollute the numerical accuracy. Based on a novel connection between the characteristic equation and the Chebyshev polynomials, we present explicit formulas for computing $V$ and $V^{-1}$, by which we prove that $mathrm{Cond}_2(V)=mathcal{O}(n^{2})$. This implies that the diagonalization process is well-conditioned and the roundoff error only increases moderately as $n$ grows and thus, compared to other direct PinT algorithms, a much larger $n$ can be used to yield satisfactory parallelism. Numerical results on parallel machine are given to support our findings, where over 60 times speedup is achieved with 256 cores.
We present a new deep learning method, dubbed FibrilNet, for tracing chromospheric fibrils in Halpha images of solar observations. Our method consists of a data pre-processing component that prepares training data from a threshold-based tool, a deep
learning model implemented as a Bayesian convolutional neural network for probabilistic image segmentation with uncertainty quantification to predict fibrils, and a post-processing component containing a fibril-fitting algorithm to determine fibril orientations. The FibrilNet tool is applied to high-resolution Halpha images from an active region (AR 12665) collected by the 1.6 m Goode Solar Telescope (GST) equipped with high-order adaptive optics at the Big Bear Solar Observatory (BBSO). We quantitatively assess the FibrilNet tool, comparing its image segmentation algorithm and fibril-fitting algorithm with those employed by the threshold-based tool. Our experimental results and major findings are summarized as follows. First, the image segmentation results (i.e., detected fibrils) of the two tools are quite similar, demonstrating the good learning capability of FibrilNet. Second, FibrilNet finds more accurate and smoother fibril orientation angles than the threshold-based tool. Third, FibrilNet is faster than the threshold-based tool and the uncertainty maps produced by FibrilNet not only provide a quantitative way to measure the confidence on each detected fibril, but also help identify fibril structures that are not detected by the threshold-based tool but are inferred through machine learning. Finally, we apply FibrilNet to full-disk Halpha images from other solar observatories and additional high-resolution Halpha images collected by BBSO/GST, demonstrating the tools usability in diverse datasets.
