Search

Your search keyword '"Liu, Shujie"' showing total 44 results
Start Over Author "Liu, Shujie" Publisher ieee
44 results on '"Liu, Shujie"'

11. Exploring WavLM on Speech Enhancement

Author

Song, Hyungchan, primary, Chen, Sanyuan, additional, Chen, Zhuo, additional, Wu, Yu, additional, Yoshioka, Takuya, additional, Tang, Min, additional, Shin, Jong Won, additional, and Liu, Shujie, additional

Published
2023
Full Text
View/download PDF

14. Optimizing Alignment of Speech and Language Latent Spaces for End-To-End Speech Recognition and Understanding

Author

Wang, Wei, Ren, Shuo, Qian, Yao, Liu, Shujie, Shi, Yu, Qian, Yanmin, and Zeng, Michael

Subjects
FOS: Computer and information sciences, Sound (cs.SD), Audio and Speech Processing (eess.AS), FOS: Electrical engineering, electronic engineering, information engineering, Computer Science - Sound, Electrical Engineering and Systems Science - Audio and Speech Processing
Abstract

The advances in attention-based encoder-decoder (AED) networks have brought great progress to end-to-end (E2E) automatic speech recognition (ASR). One way to further improve the performance of AED-based E2E ASR is to introduce an extra text encoder for leveraging extensive text data and thus capture more context-aware linguistic information. However, this approach brings a mismatch problem between the speech encoder and the text encoder due to the different units used for modeling. In this paper, we propose an embedding aligner and modality switch training to better align the speech and text latent spaces. The embedding aligner is a shared linear projection between text encoder and speech encoder trained by masked language modeling (MLM) loss and connectionist temporal classification (CTC), respectively. The modality switch training randomly swaps speech and text embeddings based on the forced alignment result to learn a joint representation space. Experimental results show that our proposed approach achieves a relative 14% to 19% word error rate (WER) reduction on Librispeech ASR task. We further verify its effectiveness on spoken language understanding (SLU), i.e., an absolute 2.5% to 2.8% F1 score improvement on SNIPS slot filling task., submitted to ICASSP 2022

Published
2022

26. Continuous Speech Separation with Conformer

Author

Chen, Sanyuan, primary, Wu, Yu, additional, Chen, Zhuo, additional, Wu, Jian, additional, Li, Jinyu, additional, Yoshioka, Takuya, additional, Wang, Chengyi, additional, Liu, Shujie, additional, and Zhou, Ming, additional

Published
2021
Full Text
View/download PDF

29. Joint Learning of Question Answering and Question Generation.

Author

Sun, Yibo, Tang, Duyu, Duan, Nan, Qin, Tao, Liu, Shujie, Yan, Zhao, Zhou, Ming, Lv, Yuanhua, Yin, Wenpeng, Feng, Xiaocheng, Qin, Bing, and Liu, Ting

Subjects
SCALABILITY, ALGORITHMS, MACHINE learning, QUESTIONING, TASK analysis
Abstract

Question answering (QA) and question generation (QG) are closely related tasks that could improve each other; however, the connection of these two tasks is not well explored in the literature. In this paper, we present two training algorithms for learning better QA and QG models through leveraging one another. The first algorithm extends Generative Adversarial Network (GAN), which selectively incorporates artificially generated instances as additional QA training data. The second algorithm is an extension of dual learning, which incorporates the probabilistic correlation of QA and QG as additional regularization in training objectives. To test the scalability of our algorithms, we conduct experiments on both document based and table based question answering tasks. Results show that both algorithms improve a QA model in terms of accuracy and QG model in terms of BLEU score. Moreover, we find that the performance of a QG model could be easily improved by a QA model via policy gradient, however, directly applying GAN that regards all the generated questions as negative instances could not improve the accuracy of the QA model. Our algorithm that selectively assigns labels to generated questions would bring a performance boost. [ABSTRACT FROM AUTHOR]

Published
2020
Full Text
View/download PDF

30. A Hierarchical Clustering Approach to Fuzzy Semantic Representation of Rare Words in Neural Machine Translation.

Author

Yang, Muyun, Liu, Shujie, Chen, Kehai, Zhang, Hongyang, Zhao, Enbo, and Zhao, Tiejun

Subjects
TRANSLATIONS, VOCABULARY, INFORMATION resources
Abstract

Rare words are usually replaced with a single $< $ $unk$ $>$ token in the current encoder–decoder style of neural machine translation, challenging the translation modeling by an obscured context. In this article, we propose to build a fuzzy semantic representation (FSR) method for rare words through a hierarchical clustering method to group rare words together, and integrate it into the encoder–decoder framework. This hierarchical structure can compensate for the semantic information in both source and target sides, and providing fuzzy context information to capture the semantic of rare words. The introduced FSR can also alleviate the data sparseness, which is the bottleneck in dealing with rare words in neural machine translation. In particular, our method is easily extended to the transformer-based neural machine translation model and learns the FSRs of all in-vocabulary words to enhance the sentence representations in addition to rare words. Our experiments on Chinese-to-English translation tasks confirm a significant improvement in the translation quality brought by the proposed method. [ABSTRACT FROM AUTHOR]

Published
2020
Full Text
View/download PDF

31. Electromagnetic Forming with Solenoid Coil

Author

Li Zhong, Wang Mei, Chen Lu, Li Junping, and Liu Shujie

Subjects
Electromagnetic field, Conducted electromagnetic interference, Engineering, business.industry, Electromagnetic coil, Electromagnetic shielding, Electrical engineering, Near and far field, Mechanics, Optical field, business, Electromagnetic pulse, Electromagnetic forming
Abstract

A loose coupling method is proposed to analyze electromagnetic forming with solenoid coil. Analysis of electromagnetic field is conducted by ANSYS/Emag, structural field is solved by explicit module ANSYS/Lsdyna, and data transfer from electromagnetic field o structural field is realized by APDL program Time-space distribution of magnetic pressure and discharge voltage, tube thickness and gap are obtained. Effect of discharge voltage, gap between coil and tube and tube thickness on electromagnetic forming are discussed respectively.

Published
2015

34. Normalized Relative RBC-Based Minimum Risk Bayesian Decision Approach for Fault Diagnosis of Industrial Process.

Author

Zheng, Ying, Mao, Simin, Liu, Shujie, Wong, David Shan-Hill, and Wang, Yan-Wei

Subjects
BAYESIAN analysis, FAULT diagnosis, MANUFACTURING processes, DENSITY, MULTIPLE correspondence analysis (Statistics)
Abstract

In the industrial process, it is helpful to take the previous fault diagnosis results into consideration during the current determination of faulty variables. In this paper, an unsupervised data-driven fault diagnosis method based on normalized relative reconstruction-based contribution (RBC) and the minimum risk Bayesian (MRB) decision theory is presented. Normalized relative RBC is used to represent the characteristic of the observation of the samples, and beta distribution is adopted to approximate the probability density distribution of the variable being faulty or normal. On the basis of the adjustable loss function, the conditional risk is obtained. The fault diagnosis method based on MRB decision is proposed, which will reduce the influence of smearing effect, improve the diagnosis rate, handle faults with a small magnitude, and identify multiple process faults. Numerical simulation examples and the Tennessee Eastman process are given to show the effectiveness and superiority of the proposed method. [ABSTRACT FROM AUTHOR]

Published
2016
Full Text
View/download PDF

44. New Depth Coding Techniques With Utilization of Corresponding Video.

Author

Liu, Shujie, Lai, PoLin, Tian, Dong, and Chen, Chang Wen

Subjects
*CODING theory, *VIDEO processing, *3-D television, *PIXELS, *RENDERING (Computer graphics), *KERNEL functions, *IMAGE quality in imaging systems
Abstract

With the development of 3D display and interactive multimedia systems, new 3D video applications, such as 3DTV and Free Viewpoint Video, are attracting significant interests. In order to enable these new applications, new data formats including captured 2D video sequences and corresponding depth maps have been proposed. Compared to conventional video frames, depth maps have very different characteristics. First, they typically consist of homogeneous areas partitioned by sharp edges representing depth discontinuities, while the sharp discontinuities play very important roles in view rendering. Second, there exists structure similarity between depth map and corresponding video, in which the edges in depth exhibit quite similar behaviors as the edges in the corresponding video. In conventional video coding techniques with transforms followed by quantization, there usually exist large artifacts along sharp edges and it costs significant more bits to represent the edges with higher accuracy. In order to suppress the coding artifacts while preserving edges, and to better compress depth information, we propose in this paper two new depth compression techniques: Trilateral Filter and Sparse Dyadic Mode. Both techniques utilize the structure similarity between depth and corresponding video while focusing on different aspects in depth compression. As a new in-loop filter, Trilateral Filter is designed to filter depth with coding artifacts based on the proximity of pixel positions, the similarity among depth samples, as well as the similarity among the collocated pixels in the video frame. While Sparse Dyadic Mode is used as an intra mode to reconstruct depth map with sparse representations of depth blocks and effective reference of edge information from video frames. With these two new coding tools, we can achieve up to about 1.5 dB gain on rendering quality as compared to depth sequences coded using MVC under the same coding rate. [ABSTRACT FROM AUTHOR]

Published
2011
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results