检索结果-内蒙古大学图书馆

Decoupled dialogue modeling and semantic parsing for multi-turn text-to-SQL

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Chen, Zhi Chen, L. Li, Hanqi Cao, Ruisheng Ma, Da Wu, Mengyue Yu, Kai X-LANCE Lab Department of Computer Science and Engineering MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai China State Key Lab of Media Convergence Production Technology and Systems Beijing China

Recently, Text-to-SQL for multi-turn dialogue has attracted great interest. Here, the user input of the current turn is parsed into the corresponding SQL query of the appropriate database, given all previous dialogue history. Current approaches mostly employ end-to-end models and consequently face two challenges. First, dialogue history modeling and Text-to- SQL parsing are implicitly combined, hence it is hard to carry out interpretable analysis and obtain targeted improvement. Second, SQL annotation of multi-turn dialogue is very expensive, leading to training data sparsity. In this paper, we propose a novel decoupled multi-turn Text-to-SQL framework, where an utterance rewrite model first explicitly solves completion of dialogue context, and then a single-turn Text-to-SQL parser follows. A dual learning approach is also proposed for the utterance rewrite model to address the data sparsity problem. Compared with end-to-end approaches, the proposed decoupled method can achieve excellent performance without any annotated in-domain data. With just a few annotated rewrite cases, the decoupled method outperforms the released state-of-the-art endto- end models on both SParC and CoSQL datasets. Copyright © 2021, The Authors. All rights reserved.

关键词： Query processing

LARGE-SCALE SELF-SUPERVISED SPEECH REPRESENTATION LEARNING FOR AUTOMATIC SPEAKER VERIFICATION

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Chen, Zhengyang Chen, Sanyuan Wu, Yu Qian, Yao Wang, Chengyi Liu, Shujie Qian, Yanmin Zeng, Michael MoE Key Lab of Artificial Intelligence AI Institute X-LANCE Lab Department of Computer Science and Engineering Shanghai Jiao Tong University China Microsoft Corporation United States Microsoft United States

The speech representations learned from large-scale unlabeled data have shown better generalizability than those from supervised learning and thus attract a lot of interest to be applied for various downstream tasks. In this paper, we explore the limits of speech representations learned by different self-supervised objectives and datasets for automatic speaker verification (ASV), especially with a well-recognized SOTA ASV model, ECAPA-TDNN [1], as a downstream model. The representations from all hidden layers of the pre-trained model are firstly averaged with learnable weights and then fed into the ECAPA-TDNN as input features. The experimental results on Voxceleb dataset show that the weighted average representation is significantly superior to FBank, a conventional handcrafted feature for ASV. Our best single system achieves 0.537%, 0.569%, and 1.180% equal error rate (EER) on the three official trials of VoxCeleb1, separately. Accordingly, the ensemble system with three pre-trained models can further improve the EER to 0.479%, 0.536% and 1.023%. Among the three evaluation trials, our best system outperforms the winner system [2] of the VoxCeleb Speaker Recognition Challenge 2021 (VoxSRC2021) on the VoxCeleb1-E trial. Copyright © 2021, The Authors. All rights reserved.

关键词： Speech recognition

Enhancing Motor Imagery EEG Signal Classification with Simplified GoogLeNet

学校读者我要写书评

暂无评论

Enhancing Motor Imagery EEG Signal Classification with Simpl...

International Winter Workshop on Brain-computer Interface (BCI)

作者： Lu Wang Junkongshuai Wang Bo Wen Wei Mu Lusheng Liu Jiaguan Han Lihua Zhang Jie Jia Xiaoyang Kang Shanghai Engineering Research Center of AI & Robotics MOE Frontiers Center for Brain Science State Key Laboratory of Medical NeurobiologyAcademy for Engineering & Technology Laboratory for Neural Interface and Brain Computer Interface Engineering Research Center of AI & Robotics Ministry of Education Institute of AI & Robotics Institute of Meta-MedicalFudan University Shanghai China Ji Hua Laboratory Guangdong Province Foshan China Yiwu Research Institute of Fudan University Chengbei Road Yiwu City Zhejiang China Department of Rehabilitation Medicine Huashan Hospital Fudan University Shanghai China Research Center for Intelligent Sensing Zhejiang Lab Hangzhou China

The brain-computer interface is a novel tool for interaction between the brain and external devices, and is an important tool at the forefront of international brain science research. Motor imagery does not require external stimulation and is one of the most widely used BCI paradigms, and is now widely used to address the treatment of neurological impairments such as paralysis, stroke, and acromegaly. Electrophysiological source imaging is a technique based on the observation of electromagnetic signals from the cerebral cortex, reflecting the electrophysiological activity of brain nerve cells on the scalp surface of the brain. The wavelet transform, on the other hand, is commonly used for the processing of non-smooth signals and aims to extract excellent signal features. This paper combines these two algorithms as feature extraction methods. In recent years, deep learning has been increasingly combined with MI-EEG-based BCI. GoogLeNet is one of the most classical convolutional neural networks and has made a splash in ImageNet image recognition. However, using GoogLeNet to train the BCI public dataset BCI IV IIa and dataset HGD has shown problems such as long training time, overfitting, and gradient disappearance. In response, this paper proposes an improved and streamlined version of GoogLeNet for training based on this dataset to achieve an overall improvement in accuracy, training time, number of parameters, and other indicators. It is useful for the realistic application of MI-based BCI systems.

关键词： Training Wavelet transforms Surface waves Scalp Pattern classification Electrophysiology Streaming media

AlphaRotate: A rotation detection benchmark using TensorFlow

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Yang, Xue Zhou, Yue Yan, Junchi Department of Computer Science and Engineering MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai200240 China Department of Electronic Engineering Shanghai Jiao Tong University Shanghai200240 China

AlphaRotate is an open-source Tensorflow benchmark for performing scalable rotation detection on various datasets. It currently provides more than 18 popular rotation detection models under a single, well-documented API designed for use by both practitioners and researchers. AlphaRotate regards high performance, robustness, sustainability and scalability as the core concept of design, and all models are covered by unit testing, continuous integration, code coverage, maintainability checks, and visual monitoring and analysis. AlphaRotate can be installed from PyPI and is released under the Apache-2.0 License. Source code is available at https://***/yangxue0827/RotationDetection. Copyright © 2021, The Authors. All rights reserved.

关键词： Convolutional neural networks

USED: Universal Speaker Extraction and Diarization

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Ao, Junyi Yıldırım, Mehmet Sinan Tao, Ruijie Ge, Meng Wang, Shuai Qian, Yanmin Li, Haizhou Shenzhen Research Institute of Big Data School of Data Science The Chinese University of Hong Kong Shenzhen518172 China Department of Electrical and Computer Engineering National University of Singapore Singapore119077 Singapore Saw Swee Hock School of Public Health National University of Singapore Singapore117549 Singapore Shenzhen Research Institute of Big Data Shenzhen518172 China Auditory Cognition and Computational Acoustics Lab Department of Computer Science and Engineering China MoE Key Laboratory of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai200240 China

Speaker extraction and diarization are two enabling techniques for real-world speech applications. Speaker extraction aims to extract a target speaker’s voice from a speech mixture, while speaker diarization demarcates speech segments by speaker, annotating ‘who spoke when’. Previous studies have typically treated the two tasks indep.ndently. In practical applications, it is more meaningful to have knowledge about ‘who spoke what and when’, which is captured by the two tasks. The two tasks share a similar objective of disentangling speakers. Speaker extraction operates in the frequency domain, whereas diarization is in the temporal domain. It is logical to believe that speaker activities obtained from speaker diarization can benefit speaker extraction, while the extracted speech offers more accurate speaker activity detection than the speech mixture. In this paper, we propose a unified model called Universal Speaker Extraction and Diarization (USED) to address output inconsistency and scenario mismatch issues. It is designed to manage speech mixtures with varying overlap ratios and variable number of speakers. We show that the USED model significantly outperforms the competitive baselines for speaker extraction and diarization tasks on LibriMix and SparseLibriMix datasets. We further validate the diarization performance on CALLHOME, a dataset based on real recordings, and experimental results indicate that our model surpasses recently proposed approaches. Copyright © 2023, The Authors. All rights reserved.

关键词： Frequency domain analysis

Fast Quaternion Product Units for Learning Disentangled Representations in SO(3)

学校读者我要写书评

暂无评论

TechRxiv

TechRxiv 2022年

作者： Qin, Shaofei Zhang, Xuan Xu, Hongteng Xu, Yi MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai China Department of Computer Science & Engineering Texas A&M University College StationTX77843 United States Gaoling School of Artificial Intelligence Renmin University of China Beijing Key Laboratory of Big Data Management and Analysis Methods Beijing China

Real-world 3D structured data like point clouds and skeletons often can be represented as data in a 3D rotation group (denoted as SO(3)). However, most existing neural networks are tailored for the data in the Euclidean space, which makes the 3D rotation data not closed under their algebraic operations and leads to sub-optimal performance in 3D-related learning tasks. To resolve the issues caused by the above mismatching between data and model, we propose a novel non-real neuron model called quaternion product unit (QPU) to represent data on 3D rotation groups. The proposed QPU leverages quaternion algebra and the law of the 3D rotation group, representing 3D rotation data as quaternions and merging them via a weighted chain of Hamilton products. We demonstrate that the QPU mathematically maintains the SO(3) structure of the 3D rotation data during the inference process and disentangles the 3D representations into "rotation-invariant" features and "rotation-equivariant" features, respectively. Moreover, we design a fast QPU to accelerate the computation of QPU. The fast QPU applies a tree-structured data indexing process, and accordingly, leverages the power of parallel computing, which reduces the computational complexity of QPU in a single thread from O(N) to O(log N). Taking the fast QPU as a basic module, we develop a series of quaternion neural networks (QNNs), including quaternion multi-layer perceptron (QMLP), quaternion message passing (QMP), and so on. In addition, we make the QNNs compatible with conventional real-valued neural networks and applicable for both skeletons and point clouds. Experiments on synthetic and real-world 3D tasks show that the QNNs based on our fast QPUs are superior to state-of-the-art real-valued models, especially in the scenarios requiring the robustness to random rotations. © 2022, CC BY.

关键词： 3D modeling

TradeFL: A Trading Mechanism for Cross-Silo Federated Learning

学校读者我要写书评

暂无评论

TradeFL: A Trading Mechanism for Cross-Silo Federated Learni...

International Conference on Distributed Computing Systems

作者： Shijing Yuan Hongtao Lv Hongze Liu Chentao Wu Song Guo Zhi Liu Hongyang Chen Jie Li Department of Computer Science and Engineering Shanghai Jiao Tong University China School of Software & Joint SDU-NTU Centre for Artificial Intelligence Research (C-FAIR) Shandong University Department of Computing The Hong Kong Polytechnic University Hong Kong The University of Electro-Communications Tokyo Japan Zhejiang Lab Research Center for Graph Computing China MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Yancheng Blockchain Research Institute China

Cross-silo federated learning (CFL) is a distributed learning paradigm that allows organizations (e.g., financial or medical entities) to train a global model on siloed data. Recent studies on mechanisms designed for CFL, however, rarely jointly consider the potential inter-organizational competition and the lack of credibility between organizations, which may discourage organizational participation. In this paper, we investigate the problem of inter-organizational competition and credibility assurance. We propose a distributed trading mechanism, called $TradeFL$ , to incentivize organizations to contribute data and computational resources through mutual trading among organizations. Technically, TradeFL characterizes the competition among organizations and compensates for their damage incurred by competition. TradeFL runs on distributed organizations and provides credibility guarantees for compensation through a customized smart contract 1 1 Illustration of the prototype: https://***/user10963.. We prove that the interaction among organizations that contribute resources to maximize personal payoffs is a weighted potential game. Then, we propose a centralized algorithm and a distributed algorithm to determine the optimal resource contribution. Simulation results and evaluations based on real-world datasets demonstrate that our scheme achieves higher social welfare, increases the amount of contributed data by up to 64%, and improves the accuracy of the global model by at most 23.2%.

关键词：

Multi-tasking dialogue comprehension with discourse parsing

学校读者我要写书评

暂无评论

arXiv 2021年

作者： He, Yuchen Zhang, Zhuosheng Zhao, Hai Department of Computer Science and Engineering Shanghai Jiao Tong University Key Laboratory of Shanghai Education Commission for Intelligent Interaction and Cognitive Engineering Shanghai Jiao Tong University MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University

Multi-party dialogue machine reading comprehension (MRC) raises an even more challenging understanding goal on dialogue with more than two involved speakers, compared with the traditional plain passage style MRC. To accurately perform the question-answering (QA) task according to such multi-party dialogue, models have to handle fundamentally different discourse relationships from common non-dialogue plain text, where discourse relations are supposed to connect two far apart utterances in a linguistics-motivated way. To further explore the role of such unusual discourse structure on the correlated QA task in terms of MRC, we propose the first multi-task model for jointly performing QA and discourse parsing (DP) on the multi-party dialogue MRC task. Our proposed model is evaluated on the latest benchmark Molweni, whose results indicate that training with complementary tasks indeed benefits not only QA task, but also DP task itself. We further find that the joint model is distinctly stronger when handling longer dialogues which again verifies the necessity of DP in the related MRC. Copyright © 2021, The Authors. All rights reserved.

关键词：

Effects of Input Neuron Mapping Coordinates in Spiking Neural Network on the Motor Imagery EEG Signals Classification

学校读者我要写书评

暂无评论

Effects of Input Neuron Mapping Coordinates in Spiking Neura...

International Winter Workshop on Brain-computer Interface (BCI)

作者： Gege Zhan Haolong Su Pengchao Wang Lan Niu Jianxiong Bin Wei Mu Xueze Zhang Haifeng Jiang Lihua Zhang Xiaoyang Kang Shanghai Engineering Research Center of AI & Robotics MOE Frontiers Center for Brain Science State Key Laboratory of Medical Neurobiology Institute of AI & Robotics Laboratory for Neural Interface and Brain Computer Interface Engineering Research Center of AI & Robotics Ministry of Education Institute of Meta-Medical Academy for Engineering & Technology Fudan University Shanghai China Ji Hua Laboratory Foshan Guangdong Province China Shanghai Mental Center Shanghai Jiao Tong University School of Medicine Shanghai China CAS Center for Excellence in Brain Science and Intelligence Technology (CEBSIT) Chinese Academy of Sciences Shanghai China Yiwu Research Institute of Fudan University Chengbei Road Yiwu City Zhejiang China Research Center for Intelligent Sensing Zhejiang Lab Hangzhou China

Spiking neural network (SNN), the third generation of the artificial neural network, uses the same computational principles as Spatio-temporal brain data (STBD) generation, so it can be used as an effective tool to learn and understand STBD. When using SNN to process STBD, mapping of input neurons is a problem to be considered. This paper investigates the influence of changing the mapping coordinates of input neurons on motor imagery EEG signal classification. The results show that the coordinate change affects the interaction between the input neuron and its neighboring neurons, and causes the classification accuracy to change, but the impact mechanism is unknown.

关键词： Neurons Pattern classification Artificial neural networks Electroencephalography Brain-computer interfaces Biological neural networks