基于重建误差的任务型对话未知意图检测

现有未知意图检测模型通常将语句映射到向量空间,并使用局部异常因子算法定义密度较小的特征点为未知意图,但经交叉熵损失训练的已知意图特征簇更加狭长,簇内的整体间距、密度和分散情况不均匀,进而增加了检测难度。针对上述问题,提出一种基于自动编码器重建误差的未知意图检测模型。在训练阶段,使用融入标签知识的联合损失函数训练已知意图分类器,使已知意图特征类间距离大且类内距离小,并利用这些特征训练一个仅能获取已知意图信息的自动编码器。在测试阶段,利用自动编码器将重建误差较大的样本视为未知意图,其余样本视为已知意图正常分类。在SNIPS数据集上的实验结果表明,在已知意图占比为25%、50%、75%时,该模型的Macro F1得分相比于表现最优的增强语义的高斯混合损失基线模型分别提升了16.93%、1.14%和2.37%,能够检测到更多的未知意图样本,同时在类别分布极不平衡的ATIS数据集上也有较好的性能表现。     

A novel task-oriented framework for dual-arm robotic assembly task

Frontiers of Mechanical Engineering - In industrial manufacturing, the deployment of dual-arm robots in assembly tasks has become a trend. However, making the dual-arm robots more intelligent in...     

Q2SM: 基于BERT的多领域任务型对话系统状态跟踪算法

基于管道的方法是目前任务型对话系统的主要构建方式,在工业界具有广泛应用,而对话状态跟踪(dialogue state tracking,DST)是任务型对话系统中的核心任务。面对传统的方法在多领域场景下表现较差的问题,该文结合语言模型预训练的最新研究成果,该文提出了一种基于BERT的对话状态跟踪算法Q2SM(query to state model)。该模型的上游使用了基于BERT的句子表征与相似度交互的槽判定模块,下游使用了一种面向对话状态跟踪任务的自定义RNN: DST-RNN。在WOZ 2.0和MultiWOZ 2.0两个数据集上的实验表明,Q2SM相比于之前的最好模型,分别在联合准确率和状态F₁值两个评价指标上提升了1.09%和2.38%。此外,模型消融实验验证了,DST-RNN相比于传统的RNN或LSTM,不仅可以提升评价指标值,还可以加快模型的收敛速度。     

Modular manufacturing

This paper discusses requirements to be satisfied by future manufacturing systems and proposes a new concept of modular manufacturing to integrate intelligent and complex machines. In large-scale systems such as manufacturing systems, modularization is indispensable for clarifying logical structure and assuring a high degree of ease of construction. The parts, products and manufacturing equipments as well as the design and operating activities themselves are all described in units called modules. A manufacturing system is constructed and operated by combining these in building-block style. The creation of this manufacturing system relies on construction and operating systems that enable design and simulation in the virtual world, and production and control in the real world, in a unified approach. Hardware modules and software modules are compiled flexibly and hierarchically to fulfil specified tasks. A system in which modular manufacturing as a concept of system integration is applied to manufacturing robots is called a modular robot system. The robots are embedded in manufacturing systems as the highest application of model-based robotics.     

Plan parsing for intended response recognition in discourse1

In a discourse the hearer must recognize the response intended by the speaker. To perform this recognition, the hearer must ascertain what plans the speaker is undertaking and how the utterances in the discourse further that plan. To do so, the hearer can parse the initial intentions (recoverable from the utterance) and recognize the plans the speaker has in mind and intends the hearer to know about. This paper reports on a theory of parsing the intentions in discourse. It also discusses the role of another aspect of discourse, discourse markers, that are valuable to intended response recognition.     

任务驱动教学法在人像数码摄影教学中的探索

在学生了解人像数码摄影概念后,通过任务驱动教学法为学生构建一个真实的人像摄影环境,让学生带着任务学习拍摄人像,在熟练掌握技能的同时,更要注重引导学生去探索发现自己周围的具体形象,练就一双善于发现事物特殊之处的慧眼,不断提高快速抓拍的能力。     

基于双层记忆网络的多领域端到端任务型对话系统

为了将知识有效地融入到对话推理,提出了一种基于双层记忆网络的多领域端到端任务型对话系统。首先,该模型用知识行的形式代替三元组的形式表示知识,提升了知识定位的性能;其次,采用了双层记忆网络结构将知识和对话历史进行分别建模,提高了模型的推理能力;最后,使用了动态编码器对多种领域的数据进行编码,提升模型的泛化能力。通过实验分析,该模型的F₁和BLEU指标在InCar和CamRest数据集上相较于对比算法均有一定的提升,验证了该模型的有效性和先进性。     

一种面向任务、可重构的主控中心设计方案

针对在大规模企业分布式网络中快速实施安全评估这一分布式安全评估问题,提出一种面向任务、可重构的主控中心方案.实验验证了该方案在快速任务生成时是可行、有效的,具有以扫描策略为组件、面向任务、可重构特性.