The construction of a case event logic graph for the judgment documentcan more intuitively retrospect the development of the case. This paperproposes a joint model of event extraction and relationship recognition for ...The construction of a case event logic graph for the judgment documentcan more intuitively retrospect the development of the case. This paperproposes a joint model of event extraction and relationship recognition for judgmentdocuments. By extracting the case information in the judgment document,a case event logic graph was constructed. The development process of the casewas shown, and a reference was provided for the analysis of the context of thecase. The experimental results show that the proposed method can extract eventsand identify the relationship between events, and the F1 value reaches 0.809. Thecase event logic graph reveals the development context of the case accurately andvividly.展开更多
This paper begins with an overview of quantum mechanics, and then recounts a relatively recent algebraic extension of the Boolean algebra of probabilistic events to “conditional events” (order pairs of events). The ...This paper begins with an overview of quantum mechanics, and then recounts a relatively recent algebraic extension of the Boolean algebra of probabilistic events to “conditional events” (order pairs of events). The main point is to show that a so-called “superposition” of two (or more) quantum events (usually with mutually inconsistent initial conditions) can be represented in this algebra of conditional events and assigned a consistent conditional probability. There is no need to imagine that a quantum particle can simultaneously straddle two inconsistent possibilities.展开更多
针对深度事件检测模型对复杂时序事件检测准确性不足和忽略了不同事件间相关性的问题,提出一种基于信号时态逻辑的深度时序事件检测算法DSTL(Deep Signal Temporal Logic)。该算法一方面引入信号时态逻辑框架,并用信号时态逻辑(STL)公...针对深度事件检测模型对复杂时序事件检测准确性不足和忽略了不同事件间相关性的问题,提出一种基于信号时态逻辑的深度时序事件检测算法DSTL(Deep Signal Temporal Logic)。该算法一方面引入信号时态逻辑框架,并用信号时态逻辑(STL)公式建模时间序列中的事件来综合考虑时间序列上事件的逻辑性和时态性;另一方面采用基于神经网络的基础分类器来检测原子事件的发生情况,并通过STL公式结构和语义来辅助检测复杂事件。另外,使用神经网络模块替代相应的逻辑连接词和时态逻辑算子,从而提供可GPU加速和梯度下降的神经网络模块。通过对6个时间序列数据集的实验,验证了该算法在时序事件检测方面的有效性,并把使用DSTL算法的模型与不使用该算法而使用多层感知机(MLP)、长短期记忆(LSTM)网络和Transformer的深度时间序列分类模型进行比较。实验结果表明,使用DSTL算法的模型在5种事件上的平均F1分数提升了约12%,其中3种跨时间点事件上的平均F1分数提升了约14%,且具备更好的可解释性。展开更多
基金This work was supported in part by the National Key R&D Program of China under Grant 2018YFC0830104.
文摘The construction of a case event logic graph for the judgment documentcan more intuitively retrospect the development of the case. This paperproposes a joint model of event extraction and relationship recognition for judgmentdocuments. By extracting the case information in the judgment document,a case event logic graph was constructed. The development process of the casewas shown, and a reference was provided for the analysis of the context of thecase. The experimental results show that the proposed method can extract eventsand identify the relationship between events, and the F1 value reaches 0.809. Thecase event logic graph reveals the development context of the case accurately andvividly.
文摘This paper begins with an overview of quantum mechanics, and then recounts a relatively recent algebraic extension of the Boolean algebra of probabilistic events to “conditional events” (order pairs of events). The main point is to show that a so-called “superposition” of two (or more) quantum events (usually with mutually inconsistent initial conditions) can be represented in this algebra of conditional events and assigned a consistent conditional probability. There is no need to imagine that a quantum particle can simultaneously straddle two inconsistent possibilities.
文摘针对深度事件检测模型对复杂时序事件检测准确性不足和忽略了不同事件间相关性的问题,提出一种基于信号时态逻辑的深度时序事件检测算法DSTL(Deep Signal Temporal Logic)。该算法一方面引入信号时态逻辑框架,并用信号时态逻辑(STL)公式建模时间序列中的事件来综合考虑时间序列上事件的逻辑性和时态性;另一方面采用基于神经网络的基础分类器来检测原子事件的发生情况,并通过STL公式结构和语义来辅助检测复杂事件。另外,使用神经网络模块替代相应的逻辑连接词和时态逻辑算子,从而提供可GPU加速和梯度下降的神经网络模块。通过对6个时间序列数据集的实验,验证了该算法在时序事件检测方面的有效性,并把使用DSTL算法的模型与不使用该算法而使用多层感知机(MLP)、长短期记忆(LSTM)网络和Transformer的深度时间序列分类模型进行比较。实验结果表明,使用DSTL算法的模型在5种事件上的平均F1分数提升了约12%,其中3种跨时间点事件上的平均F1分数提升了约14%,且具备更好的可解释性。
