The capability of error detection of patient-specific QA tools plays an important role in verifying MLC motion accuracy. The goal of this study was to investigate the capability in error detection of portal dosimetry,...The capability of error detection of patient-specific QA tools plays an important role in verifying MLC motion accuracy. The goal of this study was to investigate the capability in error detection of portal dosimetry, MapCHECK2 and MatriXX QA tools in IMRT plans. The 9 fields IMRT for 4 head and neck plans and 7 fields IMRT for 4 prostate plans were selected for the error detection of QA devices. The measurements were undertaken for the original plan and the modified plans, where the known errors were introduced for increasing and decreasing of prescribed dose (±2%, ±4% and ±6%) and position shifted in X-axis and Y-axis (±1, ±2, ±3 and ±5 mm). After measurement, the results were compared between calculated and measured values using gamma analysis at 3%/3 mm criteria. The average gamma pass for no errors introduced in head and neck plans was 96.9%, 98.6%, and 98.8%, while prostate plans presented 99.4%, 99.0%, and 99.7%, for portal dosimetry, MapCHECK2 and MatriXX system, respectively. In head and neck plan, the shifted error detections were 1 mm for portal dosimetry, 2 mm for MapCHECK2, and 3 mm for MatriXX system. In prostate plan, the shifted error detections were 2 mm for portal dosimetry, 3 mm for MapCHECK2, and 5 mm for MatriXX system. For the dose error detection, the portal dosimetry system could detect at 2% dose deviation in head and neck and 4% in prostate plans, while other two devices could detect at 4% dose deviation in both head and neck and prostate plans. Portal dosimetry shows slightly more capability to detect the error compared with MapCHECK2 and MatriXX system, especially in the complicated plan. It may be due to higher resolution of the detector;however, all three-detector types can detect various errors and can be used for patient-specific IMRT QA.展开更多
口语理解(spoken language understanding,SLU)是面向任务的对话系统的核心组成部分,旨在提取用户查询的语义框架.在对话系统中,口语理解组件(SLU)负责识别用户的请求,并创建总结用户需求的语义框架,SLU通常包括两个子任务:意图检测(int...口语理解(spoken language understanding,SLU)是面向任务的对话系统的核心组成部分,旨在提取用户查询的语义框架.在对话系统中,口语理解组件(SLU)负责识别用户的请求,并创建总结用户需求的语义框架,SLU通常包括两个子任务:意图检测(intent detection,ID)和槽位填充(slot filling,SF).意图检测是一个语义话语分类问题,在句子层面分析话语的语义;槽位填充是一个序列标注任务,在词级层面分析话语的语义.由于意图和槽之间的密切相关性,主流的工作采用联合模型来利用跨任务的共享知识.但是ID和SF是两个具有强相关性的不同任务,它们分别表征了话语的句级语义信息和词级信息,这意味着两个任务的信息是异构的,同时具有不同的粒度.提出一种用于联合意图检测和槽位填充的异构交互结构,采用自注意力和图注意力网络的联合形式充分地捕捉两个相关任务中异构信息的句级语义信息和词级信息之间的关系.不同于普通的同构结构,所提模型是一个包含不同类型节点和连接的异构图架构,因为异构图涉及更全面的信息和丰富的语义,同时可以更好地交互表征不同粒度节点之间的信息.此外,为了更好地适应槽标签的局部连续性,利用窗口机制来准确地表示词级嵌入表示.同时结合预训练模型(BERT),分析所提出模型应用预训练模型的效果.所提模型在两个公共数据集上的实验结果表明,所提模型在意图检测任务上准确率分别达到了97.98%和99.11%,在槽位填充任务上F1分数分别达到96.10%和96.11%,均优于目前主流的方法.展开更多
随着预训练语言模型在自然语言处理(NLP)任务上的应用,意图检测(ID)和槽位填充(SF)联合建模提高了口语理解的性能。现有方法大多关注意图和槽位的相互作用,忽略了差异文本序列建模对口语理解(SLU)任务的影响。因此,提出一种基于多任务...随着预训练语言模型在自然语言处理(NLP)任务上的应用,意图检测(ID)和槽位填充(SF)联合建模提高了口语理解的性能。现有方法大多关注意图和槽位的相互作用,忽略了差异文本序列建模对口语理解(SLU)任务的影响。因此,提出一种基于多任务学习的意图检测和槽位填充联合方法(IDSFML)。首先,使用随机掩盖mask策略构造差异文本,设计结合自编码器和注意力机制的神经网络(AEA)结构,为口语理解任务融入差异文本序列的特征;其次,设计相似性分布任务,使差异文本和原始文本的表征相似;最后,联合训练ID、SF和差异文本序列相似性分布三个任务。在航班旅行信息系统(ATIS)和SNIPS数据集上的实验结果表明,IDSFML与表现次优的基线方法SASGBC(Self-Attention and Slot-Gated on top of BERT with CRF)相比,槽位填充F1值分别提升了1.9和1.6个百分点,意图检测准确率分别提升了0.2和0.4个百分点,提高了口语理解任务的准确率。展开更多
为了解决工业生产设备故障领域的问答系统缺乏标注数据、意图识别槽位填充性能不足的问题,提出了一种基于Transformer的多层双向自注意编码器(bidirectional encoder representations from transformers,BERT)的联合模型。利用BERT进行...为了解决工业生产设备故障领域的问答系统缺乏标注数据、意图识别槽位填充性能不足的问题,提出了一种基于Transformer的多层双向自注意编码器(bidirectional encoder representations from transformers,BERT)的联合模型。利用BERT进行文本序列编码,并通过双向长短时记忆网络(bidirectional long short-term memory,Bi-LSTM)捕捉文本上下文语义关系。通过最大池化和致密层提取关键信息,同时使用条件随机场(conditional random field,CRF)增强模型泛化能力。构建了工业领域设备故障问答语料库,并提出了针对该领域的模型部署框架。在ATIS等公共数据集上进行实验,相对于基线模型,本文模型在句子级准确率、F 1和意图识别准确率上,分别提高4.4、2.1和0.5个百分点。研究结果有效提升了问答系统性能,为缺乏工业生产数据的问答系统领域提供了数据集和部署框架。展开更多
Traditional monitoring systems that are used in shopping malls or com-munity management,mostly use a remote control to monitor and track specific objects;therefore,it is often impossible to effectively monitor the enti...Traditional monitoring systems that are used in shopping malls or com-munity management,mostly use a remote control to monitor and track specific objects;therefore,it is often impossible to effectively monitor the entire environ-ment.Whenfinding a suspicious person,the tracked object cannot be locked in time for tracking.This research replaces the traditionalfixed-point monitor with the intelligent drone and combines the image processing technology and automatic judgment for the movements of the monitored person.This intelligent system can effectively improve the shortcomings of low efficiency and high cost of the traditional monitor system.In this article,we proposed a TIMT(The Intel-ligent Monitoring and Tracking)algorithm which can make the drone have smart surveillance and tracking capabilities.It combined with Artificial Intelligent(AI)face recognition technology and the OpenPose which is able to monitor the phy-sical movements of multiple people in real time to analyze the meaning of human body movements and to track the monitored intelligently through the remote con-trol interface of the drone.This system is highly agile and could be adjusted immediately to any angle and screen that we monitor.Therefore,the system couldfind abnormal conditions immediately and track and monitor them automatically.That is the system can immediately detect when someone invades the home or community,and the drone can automatically track the intruder to achieve that the two significant shortcomings of the traditional monitor will be improved.Experimental results show that the intelligent monitoring and tracking drone sys-tem has an excellent performance,which not only dramatically reduces the num-ber of monitors and the required equipment but also achieves perfect monitoring and tracking.展开更多
Few-shot intent detection is a practical challenge task,because new intents are frequently emerging and collecting large-scale data for them could be costly.Meta-learning,a promising technique for leveraging data from...Few-shot intent detection is a practical challenge task,because new intents are frequently emerging and collecting large-scale data for them could be costly.Meta-learning,a promising technique for leveraging data from previous tasks to enable efficient learning of new tasks,has been a popular way to tackle this problem.However,the existing meta-learning models have been evidenced to be overfitting when the meta-training tasks are insufficient.To overcome this challenge,we present a novel self-supervised task augmentation with meta-learning framework,namely STAM.Firstly,we introduce the task augmentation,which explores two different strategies and combines them to extend meta-training tasks.Secondly,we devise two auxiliary losses for integrating self-supervised learning into meta-learning to learn more generalizable and transferable features.Experimental results show that STAM can achieve consistent and considerable performance improvement to existing state-of-the-art methods on four datasets.展开更多
文摘The capability of error detection of patient-specific QA tools plays an important role in verifying MLC motion accuracy. The goal of this study was to investigate the capability in error detection of portal dosimetry, MapCHECK2 and MatriXX QA tools in IMRT plans. The 9 fields IMRT for 4 head and neck plans and 7 fields IMRT for 4 prostate plans were selected for the error detection of QA devices. The measurements were undertaken for the original plan and the modified plans, where the known errors were introduced for increasing and decreasing of prescribed dose (±2%, ±4% and ±6%) and position shifted in X-axis and Y-axis (±1, ±2, ±3 and ±5 mm). After measurement, the results were compared between calculated and measured values using gamma analysis at 3%/3 mm criteria. The average gamma pass for no errors introduced in head and neck plans was 96.9%, 98.6%, and 98.8%, while prostate plans presented 99.4%, 99.0%, and 99.7%, for portal dosimetry, MapCHECK2 and MatriXX system, respectively. In head and neck plan, the shifted error detections were 1 mm for portal dosimetry, 2 mm for MapCHECK2, and 3 mm for MatriXX system. In prostate plan, the shifted error detections were 2 mm for portal dosimetry, 3 mm for MapCHECK2, and 5 mm for MatriXX system. For the dose error detection, the portal dosimetry system could detect at 2% dose deviation in head and neck and 4% in prostate plans, while other two devices could detect at 4% dose deviation in both head and neck and prostate plans. Portal dosimetry shows slightly more capability to detect the error compared with MapCHECK2 and MatriXX system, especially in the complicated plan. It may be due to higher resolution of the detector;however, all three-detector types can detect various errors and can be used for patient-specific IMRT QA.
文摘口语理解(spoken language understanding,SLU)是面向任务的对话系统的核心组成部分,旨在提取用户查询的语义框架.在对话系统中,口语理解组件(SLU)负责识别用户的请求,并创建总结用户需求的语义框架,SLU通常包括两个子任务:意图检测(intent detection,ID)和槽位填充(slot filling,SF).意图检测是一个语义话语分类问题,在句子层面分析话语的语义;槽位填充是一个序列标注任务,在词级层面分析话语的语义.由于意图和槽之间的密切相关性,主流的工作采用联合模型来利用跨任务的共享知识.但是ID和SF是两个具有强相关性的不同任务,它们分别表征了话语的句级语义信息和词级信息,这意味着两个任务的信息是异构的,同时具有不同的粒度.提出一种用于联合意图检测和槽位填充的异构交互结构,采用自注意力和图注意力网络的联合形式充分地捕捉两个相关任务中异构信息的句级语义信息和词级信息之间的关系.不同于普通的同构结构,所提模型是一个包含不同类型节点和连接的异构图架构,因为异构图涉及更全面的信息和丰富的语义,同时可以更好地交互表征不同粒度节点之间的信息.此外,为了更好地适应槽标签的局部连续性,利用窗口机制来准确地表示词级嵌入表示.同时结合预训练模型(BERT),分析所提出模型应用预训练模型的效果.所提模型在两个公共数据集上的实验结果表明,所提模型在意图检测任务上准确率分别达到了97.98%和99.11%,在槽位填充任务上F1分数分别达到96.10%和96.11%,均优于目前主流的方法.
文摘随着预训练语言模型在自然语言处理(NLP)任务上的应用,意图检测(ID)和槽位填充(SF)联合建模提高了口语理解的性能。现有方法大多关注意图和槽位的相互作用,忽略了差异文本序列建模对口语理解(SLU)任务的影响。因此,提出一种基于多任务学习的意图检测和槽位填充联合方法(IDSFML)。首先,使用随机掩盖mask策略构造差异文本,设计结合自编码器和注意力机制的神经网络(AEA)结构,为口语理解任务融入差异文本序列的特征;其次,设计相似性分布任务,使差异文本和原始文本的表征相似;最后,联合训练ID、SF和差异文本序列相似性分布三个任务。在航班旅行信息系统(ATIS)和SNIPS数据集上的实验结果表明,IDSFML与表现次优的基线方法SASGBC(Self-Attention and Slot-Gated on top of BERT with CRF)相比,槽位填充F1值分别提升了1.9和1.6个百分点,意图检测准确率分别提升了0.2和0.4个百分点,提高了口语理解任务的准确率。
文摘为了解决工业生产设备故障领域的问答系统缺乏标注数据、意图识别槽位填充性能不足的问题,提出了一种基于Transformer的多层双向自注意编码器(bidirectional encoder representations from transformers,BERT)的联合模型。利用BERT进行文本序列编码,并通过双向长短时记忆网络(bidirectional long short-term memory,Bi-LSTM)捕捉文本上下文语义关系。通过最大池化和致密层提取关键信息,同时使用条件随机场(conditional random field,CRF)增强模型泛化能力。构建了工业领域设备故障问答语料库,并提出了针对该领域的模型部署框架。在ATIS等公共数据集上进行实验,相对于基线模型,本文模型在句子级准确率、F 1和意图识别准确率上,分别提高4.4、2.1和0.5个百分点。研究结果有效提升了问答系统性能,为缺乏工业生产数据的问答系统领域提供了数据集和部署框架。
文摘Traditional monitoring systems that are used in shopping malls or com-munity management,mostly use a remote control to monitor and track specific objects;therefore,it is often impossible to effectively monitor the entire environ-ment.Whenfinding a suspicious person,the tracked object cannot be locked in time for tracking.This research replaces the traditionalfixed-point monitor with the intelligent drone and combines the image processing technology and automatic judgment for the movements of the monitored person.This intelligent system can effectively improve the shortcomings of low efficiency and high cost of the traditional monitor system.In this article,we proposed a TIMT(The Intel-ligent Monitoring and Tracking)algorithm which can make the drone have smart surveillance and tracking capabilities.It combined with Artificial Intelligent(AI)face recognition technology and the OpenPose which is able to monitor the phy-sical movements of multiple people in real time to analyze the meaning of human body movements and to track the monitored intelligently through the remote con-trol interface of the drone.This system is highly agile and could be adjusted immediately to any angle and screen that we monitor.Therefore,the system couldfind abnormal conditions immediately and track and monitor them automatically.That is the system can immediately detect when someone invades the home or community,and the drone can automatically track the intruder to achieve that the two significant shortcomings of the traditional monitor will be improved.Experimental results show that the intelligent monitoring and tracking drone sys-tem has an excellent performance,which not only dramatically reduces the num-ber of monitors and the required equipment but also achieves perfect monitoring and tracking.
基金the National Natural Science Foundation of China under Grant Nos.61936012 and 61976114。
文摘Few-shot intent detection is a practical challenge task,because new intents are frequently emerging and collecting large-scale data for them could be costly.Meta-learning,a promising technique for leveraging data from previous tasks to enable efficient learning of new tasks,has been a popular way to tackle this problem.However,the existing meta-learning models have been evidenced to be overfitting when the meta-training tasks are insufficient.To overcome this challenge,we present a novel self-supervised task augmentation with meta-learning framework,namely STAM.Firstly,we introduce the task augmentation,which explores two different strategies and combines them to extend meta-training tasks.Secondly,we devise two auxiliary losses for integrating self-supervised learning into meta-learning to learn more generalizable and transferable features.Experimental results show that STAM can achieve consistent and considerable performance improvement to existing state-of-the-art methods on four datasets.