【摘 要】
：
中文命名实体识别(NER)任务是信息抽取领域内的一个子任务,其任务目标是给定一段非结构文本后,从句子中寻找、识别和分类相关实体,例如人名、地名和机构名称.中文命名实体识别是一个自然语言处理(NLP)领域的基本任务,在许多下游NLP任务中,包括信息检索、关系抽取和问答系统中扮演着重要角色.全面回顾了现有的基于神经网络的单词-字符晶格结构的中文NER模型.首先介绍了中文NER相比英语NER难度更大,存在着中文文本相关实体边界难以确定和中文语法结构复杂等难点及挑战.然后调研了在不同神经网络架构下(RNN、CNN
【机 构】
：
国防科技大学 计算机学院,长沙 410073
论文部分内容阅读
中文命名实体识别(NER)任务是信息抽取领域内的一个子任务,其任务目标是给定一段非结构文本后,从句子中寻找、识别和分类相关实体,例如人名、地名和机构名称.中文命名实体识别是一个自然语言处理(NLP)领域的基本任务,在许多下游NLP任务中,包括信息检索、关系抽取和问答系统中扮演着重要角色.全面回顾了现有的基于神经网络的单词-字符晶格结构的中文NER模型.首先介绍了中文NER相比英语NER难度更大,存在着中文文本相关实体边界难以确定和中文语法结构复杂等难点及挑战.然后调研了在不同神经网络架构下(RNN、CNN、GNN和Transformer)最具代表性的晶格结构的中文NER模型.由于单词序列信息可以给基于字符的序列学习更多边界信息,为了显式地利用每个字符所相关的词汇信息,过去的这些工作提出通过词-字符晶格结构将单词信息整合到字符序列中.这些在中文NER任务上基于神经网络的单词-字符晶格结构的性能要明显优于基于单词或基于字符的方法.最后介绍了中文NER的数据集及评价标准.
其他文献
利用关系数据进行股价预测的方法最近已经被提出,但目前还没有找到一种有效的方法可以有选择地聚合不同类型的关系数据去预测股价.提出一种改进的多层节点图注意力网络(FHAN)模型,该方法融合Fraudar算法,提供了一种对多个对象关系之间看问题的视角.模型把公司看做节点,把交互看成边,选择性地聚合不同关系类型的信息,并将这些信息添加到每个公司的节点表示中,添加了信息的节点表示被输入到特定任务层自动选择信息,实验结果表明,该方法比目前流行的神经网络算法在股价预测的效果上更准确,实验选取不同神经网络算法做对比,在最
As a continuation of previous years\'special section on software systems,this special section encourages and promotes research to address challenges from the perspective of software systems.The goal of this special section is to present state-of-the-art
Many applications need to meet diverse requirements of a large-scale distributed user group.That challenges the current requirements engineering techniques.Crowd-based requirements engineering was proposed as an umbrella term for dealing with the requirem
Programmable logic controllers(PLCs)play a critical role in many industrial control systems,yet face in-creasingly serious cyber threats.In this paper,we propose a novel PLC-compatible software-based defense mechanism,called Heterogeneous Redundant Proact
Allocation,dereferencing,and freeing of memory data in kernels are coherently linked.There widely exist real cases where the correctness of memory is compromised.This incorrectness in kernel memory brings about significant security issues,e.g.,information
A quantum circuit is a computational unit that transforms an input quantum state to an output state.A natural way to reason about its behavior is to compute explicitly the unitary matrix implemented by it.However,when the number of qubits increases,the ma
Contextual refinement is a compositional approach to compositional verification of concurrent objects.There has been much work designing program logics to prove the contextual refinement between the object implementation and its abstract specification.How
The Linux kernel adopts a large number of security checks to prevent security-sensitive operations from being executed under unsafe conditions.If a security-sensitive operation is unchecked,a missing-check issue arises.Missing check is a class of severe b
Graph neural networks(GNNs) have shown great power in learning on graphs.However,it is still a challenge for GNNs to model information faraway from the source node.The ability to preserve global information can enhance graph representation and hence impro
Dialogue state tracking(DST)leverages dialogue information to predict dialogues states which are generally represented as slot-value pairs.However,previous work usually has limitations to efficiently predict values due to the lack of a powerful strategy f