Spoken dialogue systems are an active research field with wide applications. But the differences in the Chinese spoken dialogue system are not as distinct as that of English. In Chinese spoken dialogues, there are man...Spoken dialogue systems are an active research field with wide applications. But the differences in the Chinese spoken dialogue system are not as distinct as that of English. In Chinese spoken dialogues, there are many language phenomena. Firstly, most utterances are ill-formed. Secondly, ellipsis, anaphora and negation are also widely used in Chinese spoken dialogue. Determining how to extract semantic information from incomplete sentences and resolve negation, anaphora and ellipsis is crucial. SHTQS (Shanghai Transportation Query System) is an intelligent telephone-based spoken dialogue system providing information about the best route between any two sites in Shanghai. After a brief description of the system, the natural language processing is emphasized. Speech recognition sentences unavoidably contain errors. In language sequence processing procedures, these errors can be easily passed to the later parts and take on a ripple effect. To detect and recover these from errors as early as possible, language-processing strategies are specially considered. For errors resulting from divided words in speech recognition, segmentation and POS Tagging approaches that can rectify these errors are designed. Since most of the inquiry utterances are ill-formed and negation, anaphora and ellipsis are common language phenomena, the language understanding must be adequately adaptive. So, a partial syntactic parsing scheme is adopted and a chart algorithm is used. The parser is based on unification grammar. The semantic frame that extracts from the best arc set of the chart is used to represent the meaning of sentences. The negation, anaphora and ellipsis are also analyzed and corresponding processing approaches are presented. The accuracy of the language processing part is 88.39% and the testing result shows that the language processing strategies are rational and effective.展开更多
SHTQS is an intelligent telephone-besed spoken dialyze system providing the infomation about the best route between two sites in Shanghai. Instead of separated parts of speech decoding and language parsing, a close co...SHTQS is an intelligent telephone-besed spoken dialyze system providing the infomation about the best route between two sites in Shanghai. Instead of separated parts of speech decoding and language parsing, a close cool,ration is carded out in SHTQS by integrating automatic speech recognizer （AS,R）, language understanding, dialogue management and speech generatot. In such a way, the erroneous analysis and uncertainty happening in the preceding stages would be recovered and determined acourately with high-level knowledge, Moreover, instead of shallow word-level analysis or simply keyword or key phrase matching, a deeper analysis is performed in our system by integrating a robust parser and a semantic interpreter. The robust parser is particularly important for spontanecos speech inputs because most of the inquiry sentences/phrases are ill-formed. In addition, in designinga mixed-initiative dialogue system, understanding users＇ inquiries is essential; however, simply matching keywords and/or key phrases can hardly achieve this. Therefore, a semantic interpreter is incorporated in oar system. The performnce of is also evaluated. The dialogue efficiency is 4.4 sentences per query on an average and the case precision rate of language understanding module is up to 81%. The results are satisfactory.展开更多
ChatGPT引发了新一轮的科技革命,使得对话系统成为研究热点。口语理解(Spoken Language Understanding,SLU)作为任务型对话系统的第一部分,对系统整体的表现具有重要影响。在最近几年中,得益于大规模语言模型的成功,口语理解任务取得了...ChatGPT引发了新一轮的科技革命,使得对话系统成为研究热点。口语理解(Spoken Language Understanding,SLU)作为任务型对话系统的第一部分,对系统整体的表现具有重要影响。在最近几年中,得益于大规模语言模型的成功,口语理解任务取得了较大的发展。然而,现有工作大多基于书面语数据集完成,无法很好地应对真实口语场景。为此,该文面向与书面语相对的口语,重点关注医疗领域这一应用场景,对现有的医疗领域对话系统口语理解任务进行综述。具体地,该文阐述了医疗口语理解任务的难点与挑战,并从数据集、算法和应用的层面梳理了医疗口语理解的研究现状及不足之处。最后,该文结合生成式大模型的最新进展,给出了医疗口语理解问题新的研究方向。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
构建了基于BERT的双向连接模式BERT-based Bi-directional Association Model(BBAM)以实现在意图识别和槽位填充之间建立双向关系的目标,来实现意图识别与槽位填充的双向关联,融合两个任务的上下文信息,对意图识别与槽位填充两个任务之...构建了基于BERT的双向连接模式BERT-based Bi-directional Association Model(BBAM)以实现在意图识别和槽位填充之间建立双向关系的目标,来实现意图识别与槽位填充的双向关联,融合两个任务的上下文信息,对意图识别与槽位填充两个任务之间的联系进行深度挖掘,从而优化问句理解的整体性能.为了验证模型在旅游领域中的实用性和有效性,通过远程监督和人工校验构建了旅游领域问句数据集TFQD(Tourism Field Question Dataset),BBAM模型在此数据集上的槽填充任务F 1值得分为95.21%,意图分类准确率(A)为96.71%,整体识别准确率(A_(sentence))高达89.62%,显著优于多种基准模型.所提出的模型在ATIS和Snips两个公开数据集上与主流联合模型进行对比实验后,结果表明其具备一定的泛化能力.展开更多
文摘Spoken dialogue systems are an active research field with wide applications. But the differences in the Chinese spoken dialogue system are not as distinct as that of English. In Chinese spoken dialogues, there are many language phenomena. Firstly, most utterances are ill-formed. Secondly, ellipsis, anaphora and negation are also widely used in Chinese spoken dialogue. Determining how to extract semantic information from incomplete sentences and resolve negation, anaphora and ellipsis is crucial. SHTQS (Shanghai Transportation Query System) is an intelligent telephone-based spoken dialogue system providing information about the best route between any two sites in Shanghai. After a brief description of the system, the natural language processing is emphasized. Speech recognition sentences unavoidably contain errors. In language sequence processing procedures, these errors can be easily passed to the later parts and take on a ripple effect. To detect and recover these from errors as early as possible, language-processing strategies are specially considered. For errors resulting from divided words in speech recognition, segmentation and POS Tagging approaches that can rectify these errors are designed. Since most of the inquiry utterances are ill-formed and negation, anaphora and ellipsis are common language phenomena, the language understanding must be adequately adaptive. So, a partial syntactic parsing scheme is adopted and a chart algorithm is used. The parser is based on unification grammar. The semantic frame that extracts from the best arc set of the chart is used to represent the meaning of sentences. The negation, anaphora and ellipsis are also analyzed and corresponding processing approaches are presented. The accuracy of the language processing part is 88.39% and the testing result shows that the language processing strategies are rational and effective.
文摘SHTQS is an intelligent telephone-besed spoken dialyze system providing the infomation about the best route between two sites in Shanghai. Instead of separated parts of speech decoding and language parsing, a close cool,ration is carded out in SHTQS by integrating automatic speech recognizer （AS,R）, language understanding, dialogue management and speech generatot. In such a way, the erroneous analysis and uncertainty happening in the preceding stages would be recovered and determined acourately with high-level knowledge, Moreover, instead of shallow word-level analysis or simply keyword or key phrase matching, a deeper analysis is performed in our system by integrating a robust parser and a semantic interpreter. The robust parser is particularly important for spontanecos speech inputs because most of the inquiry sentences/phrases are ill-formed. In addition, in designinga mixed-initiative dialogue system, understanding users＇ inquiries is essential; however, simply matching keywords and/or key phrases can hardly achieve this. Therefore, a semantic interpreter is incorporated in oar system. The performnce of is also evaluated. The dialogue efficiency is 4.4 sentences per query on an average and the case precision rate of language understanding module is up to 81%. The results are satisfactory.
文摘ChatGPT引发了新一轮的科技革命,使得对话系统成为研究热点。口语理解(Spoken Language Understanding,SLU)作为任务型对话系统的第一部分,对系统整体的表现具有重要影响。在最近几年中,得益于大规模语言模型的成功,口语理解任务取得了较大的发展。然而,现有工作大多基于书面语数据集完成,无法很好地应对真实口语场景。为此,该文面向与书面语相对的口语,重点关注医疗领域这一应用场景,对现有的医疗领域对话系统口语理解任务进行综述。具体地,该文阐述了医疗口语理解任务的难点与挑战,并从数据集、算法和应用的层面梳理了医疗口语理解的研究现状及不足之处。最后,该文结合生成式大模型的最新进展,给出了医疗口语理解问题新的研究方向。
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。
文摘构建了基于BERT的双向连接模式BERT-based Bi-directional Association Model(BBAM)以实现在意图识别和槽位填充之间建立双向关系的目标,来实现意图识别与槽位填充的双向关联,融合两个任务的上下文信息,对意图识别与槽位填充两个任务之间的联系进行深度挖掘,从而优化问句理解的整体性能.为了验证模型在旅游领域中的实用性和有效性,通过远程监督和人工校验构建了旅游领域问句数据集TFQD(Tourism Field Question Dataset),BBAM模型在此数据集上的槽填充任务F 1值得分为95.21%,意图分类准确率(A)为96.71%,整体识别准确率(A_(sentence))高达89.62%,显著优于多种基准模型.所提出的模型在ATIS和Snips两个公开数据集上与主流联合模型进行对比实验后,结果表明其具备一定的泛化能力.
