期刊文献+
共找到1,064篇文章
< 1 2 54 >
每页显示 20 50 100
Enhancing Relational Triple Extraction in Specific Domains:Semantic Enhancement and Synergy of Large Language Models and Small Pre-Trained Language Models
1
作者 Jiakai Li Jianpeng Hu Geng Zhang 《Computers, Materials & Continua》 SCIE EI 2024年第5期2481-2503,共23页
In the process of constructing domain-specific knowledge graphs,the task of relational triple extraction plays a critical role in transforming unstructured text into structured information.Existing relational triple e... In the process of constructing domain-specific knowledge graphs,the task of relational triple extraction plays a critical role in transforming unstructured text into structured information.Existing relational triple extraction models facemultiple challenges when processing domain-specific data,including insufficient utilization of semantic interaction information between entities and relations,difficulties in handling challenging samples,and the scarcity of domain-specific datasets.To address these issues,our study introduces three innovative components:Relation semantic enhancement,data augmentation,and a voting strategy,all designed to significantly improve the model’s performance in tackling domain-specific relational triple extraction tasks.We first propose an innovative attention interaction module.This method significantly enhances the semantic interaction capabilities between entities and relations by integrating semantic information fromrelation labels.Second,we propose a voting strategy that effectively combines the strengths of large languagemodels(LLMs)and fine-tuned small pre-trained language models(SLMs)to reevaluate challenging samples,thereby improving the model’s adaptability in specific domains.Additionally,we explore the use of LLMs for data augmentation,aiming to generate domain-specific datasets to alleviate the scarcity of domain data.Experiments conducted on three domain-specific datasets demonstrate that our model outperforms existing comparative models in several aspects,with F1 scores exceeding the State of the Art models by 2%,1.6%,and 0.6%,respectively,validating the effectiveness and generalizability of our approach. 展开更多
关键词 Relational triple extraction semantic interaction large language models data augmentation specific domains
下载PDF
Mechatronic Modeling and Domain Transformation of Multi-physics Systems
2
作者 Clarence W.DE SILVA 《Instrumentation》 2021年第1期14-28,共15页
The enhanced definition of Mechatronics involves the four underlying characteristics of integrated,unified,unique,and systematic approaches.In this realm,Mechatronics is not limited to electro-mechanical systems,in th... The enhanced definition of Mechatronics involves the four underlying characteristics of integrated,unified,unique,and systematic approaches.In this realm,Mechatronics is not limited to electro-mechanical systems,in the multi-physics sense,but involves other physical domains such as fluid and thermal.This paper summarizes the mechatronic approach to modeling.Linear graphs facilitate the development of state-space models of mechatronic systems,through this approach.The use of linear graphs in mechatronic modeling is outlined and an illustrative example of sound system modeling is given.Both time-domain and frequency-domain approaches are presented for the use of linear graphs.A mechatronic model of a multi-physics system may be simplified by converting all the physical domains into an equivalent single-domain system that is entirely in the output domain of the system.This approach of converting(transforming)physical domains is presented.An illustrative example of a pressure-controlled hydraulic actuator system that operates a mechanical load is given. 展开更多
关键词 Mechatronic modeling Multi-physics Systems Integrated unified Unique and Systematic Approach Linear Graphs Physical domain Conversion/Transformation
下载PDF
Capability requirements modeling and verification based on fuzzy ontology 被引量:4
3
作者 Qingchao Dong Zhixue Wang Weixing Zhu Hongyue He 《Journal of Systems Engineering and Electronics》 SCIE EI CSCD 2012年第1期78-87,共10页
The capability requirements of the command, control, communication, computing, intelligence, surveillance, reconnaissance (C41SR) systems are full of uncertain and vague information, which makes it difficult to mode... The capability requirements of the command, control, communication, computing, intelligence, surveillance, reconnaissance (C41SR) systems are full of uncertain and vague information, which makes it difficult to model the C41SR architecture. The paper presents an approach to modeling the capability requirements with the fuzzy unified modeling language (UML) and building domain ontologies with fuzzy description logic (DL). The UML modeling constructs are extended according to the meta model of Depart- ment of Defense Architecture Framework to improve their domain applicability, the fuzzy modeling mechanism is introduced to model the fuzzy efficiency features of capabilities, and the capability requirement models are converted into ontologies formalized in fuzzy DL so that the model consistency and reasonability can be checked with a DL reasoning system. Finally, a case study of C41SR capability requirements model checking is provided to demonstrate the availability and applicability of the method. 展开更多
关键词 fuzzy ontology fuzzy unified modeling language (UML) fuzzy description logic (DL) model checking.
下载PDF
Formalization and Verification of Business Process Modeling Based on UML and Petri Nets 被引量:1
4
作者 颜志军 甘仞初 《Journal of Beijing Institute of Technology》 EI CAS 2005年第2期212-216,共5页
In order to provide a quantitative analysis and verification method for activity diagrams based business process modeling, a formal definition of activity diagrams is introduced. And the basic requirements for activit... In order to provide a quantitative analysis and verification method for activity diagrams based business process modeling, a formal definition of activity diagrams is introduced. And the basic requirements for activity diagrams based business process models are proposed. Furthermore, the standardized transformation technique between business process models and basic Petri nets is presented and the analysis method for the soundness and well-structured properties of business processes is introduced. 展开更多
关键词 business process modeling unified modeling language(UML) Petri nets activity diagram
下载PDF
A UML profile for framework modeling 被引量:1
5
作者 徐小良 汪乐宇 周泓 《Journal of Zhejiang University Science》 CSCD 2004年第1期92-98,共7页
The current standard Unified Modeling Language(UML) could not model framework flexibility and extendibility adequately due to lack of appropriate constructs to distinguish framework hot-spots from kernel elements. A n... The current standard Unified Modeling Language(UML) could not model framework flexibility and extendibility adequately due to lack of appropriate constructs to distinguish framework hot-spots from kernel elements. A new UML profile that may customize UML for framework modeling was presented using the extension mechanisms of UML, providing a group of UML extensions to meet the needs of framework modeling. In this profile, the extended class diagrams and sequence diagrams were defined to straightforwardly identify the hot-spots and describe their instantiation restrictions. A transformation model based on design patterns was also put forward, such that the profile based framework design diagrams could be automatically mapped to the corresponding implementation diagrams. It was proved that the presented profile makes framework modeling more straightforwardly and therefore easier to understand and instantiate. 展开更多
关键词 Object oriented frameworks unified modeling language(UML) UML profile Hot spots Design patterns
下载PDF
Spatial data modeling for coalfield geological environment
6
作者 JIA Bei SU Qiao-mei LIU Chen LI Hui-juan 《Journal of Coal Science & Engineering(China)》 2010年第3期300-305,共6页
Presented a study on the design and implementation of spatial data modelingand application in the spatial data organization and management of a coalfield geologicalenvironment database.Based on analysis of a number of... Presented a study on the design and implementation of spatial data modelingand application in the spatial data organization and management of a coalfield geologicalenvironment database.Based on analysis of a number of existing data models and takinginto account the unique data structure and characteristic, methodology and key techniquesin the object-oriented spatial data modeling were proposed for the coalfield geological environment.The model building process was developed using object-oriented technologyand the Unified Modeling Language (UML) on the platform of ESRI geodatabase datamodels.A case study of spatial data modeling in UML was presented with successful implementationin the spatial database of the coalfield geological environment.The modelbuilding and implementation provided an effective way of representing the complexity andspecificity of coalfield geological environment spatial data and an integrated managementof spatial and property data. 展开更多
关键词 spatial data model OBJECT-ORIENTED unified modeling language (UML) coal- field geological environment
下载PDF
Modeling and OLAP Cubes for Database of Ground and Municipal Water Supply
7
作者 Taskeen Zaidi Annapurna Singh Vipin Saxena 《Computational Water, Energy, and Environmental Engineering》 2013年第3期77-82,共6页
Modeling plays an important role for the solution of the complex research problems. When the database became large and complex then it is necessary to create a unified model for getting the desired information in the ... Modeling plays an important role for the solution of the complex research problems. When the database became large and complex then it is necessary to create a unified model for getting the desired information in the minimum time and to implement the model in a better way. The present paper deals with the modeling for searching of the desired information from a large database by storing the data inside the three dimensional data cubes. A sample case study is considered as a real data related to the ground water and municipal water supply, which contains the data from the various localities of a city. For the demonstration purpose, a sample size is taken as nine but when it becomes very large for number of localities of different cities then it is necessary to store the data inside data cubes. A well known object-oriented Unified Modeling Language (UML) is used to create Unified class and state models. For verification purpose, sample queries are also performed and corresponding results are depicted. 展开更多
关键词 modeling DATABASE OBJECT-ORIENTED unified modeling language OLAP Data CUBES Water Supply
下载PDF
An Embedded Software Modeling and Process by Using Aspect-Oriented Approach
8
作者 Yong-Yi FanJiang Jong-Yih Kuo Shang-Pin Ma 《Journal of Software Engineering and Applications》 2011年第2期106-122,共17页
In recent years, mobile devices have become widespread and refined, and they have offered increased convenience in human life. For these reasons, a variety of embedded systems have been designed. Therefore, improving ... In recent years, mobile devices have become widespread and refined, and they have offered increased convenience in human life. For these reasons, a variety of embedded systems have been designed. Therefore, improving methods for developing of embedded software systematically has become an important issue. Platform-based design is one example of an embedded-system design method that can reduce the design cost via improving a design’s abstraction level. However, platform-based design lacks precise definitions for platforms and design processes. This paper provides an approach that combines the aspects and platform-based design methods for developing embedded software. The approach is built on platform-based design methodology and uses the separating of concerns (SoC) concept to define the aspects and to reduce the crosscutting concerns in embedded system modeling. For aspect issues, we use the extended UML notation with aspects to describe both the static structure and the dynamic structure of the embedded system. We used an example of a digital photo frame system to demonstrate our approach. 展开更多
关键词 Platform-Based Design ASPECT-ORIENTED unified modeling language EMBEDDED SOFTWARE
下载PDF
MODELING OF FMS BASED ON UML AND OPNS
9
作者 Gao Meimei Wu Zhiming (Department of Automation, Shanghai Jiaotong University) 《Chinese Journal of Mechanical Engineering》 SCIE EI CAS CSCD 2000年第2期90-95,共6页
As the main component of computer integrated manufacturing system (CIMS), flexible manufacturing system (FMS) should be an open system with reusability and extenchaility. Moreover, as FMS is a complex asynchronous con... As the main component of computer integrated manufacturing system (CIMS), flexible manufacturing system (FMS) should be an open system with reusability and extenchaility. Moreover, as FMS is a complex asynchronous concurrent system, its model also should have the abilities to express the concurrency in the system and to analyze the behavior of the system. It is difficult to use any one method to model such a complex system as FMS. A modeling method using Object-oriented modeling language-unified modeling language (UML) and object-Oriented Petri nets (OPNs) is proposed. Class diagram in UML is used to represent the static relations among the objects in FMS. OPNs are used to model the dynamic behavior of the objects and conduct performance analysis. OPNs also can be used to identify the attributes and operations of the objects. The model can describe the system integrally and can be used to design FMS control software naturally. 展开更多
关键词 Flexible manufacturing system modeling Object-oriented model unified modeling language Object-oriented Pert nets (Opens)
下载PDF
基于MBSE的作战能力跨域描述元建模方法
10
作者 朱刚 郑建成 +2 位作者 李志淮 常春贺 刘华 《火力与指挥控制》 CSCD 北大核心 2024年第8期97-104,共8页
针对作战能力跨域描述存在语义二义性、不利于交流等问题,提出一种基于模型系统工程的元建模方法。抽象出作战能力元模型并给出形式化定义;从能力、作战和信息视角扩展作战能力元模型定义作战能力描述语言;通过UML Profile技术实现作战... 针对作战能力跨域描述存在语义二义性、不利于交流等问题,提出一种基于模型系统工程的元建模方法。抽象出作战能力元模型并给出形式化定义;从能力、作战和信息视角扩展作战能力元模型定义作战能力描述语言;通过UML Profile技术实现作战能力描述语言建模工具;借鉴面向对象分析方法思路提出作战能力描述语言建模方法。以美军“分布式防御”概念为例对提出方法进行实例验证。结果表明,提出的方法易于理解、可靠性和可扩展性强,有效消除语义二义性的同时满足多视角跨域描述作战能力的需求。 展开更多
关键词 元模型 作战能力描述语言 统一建模语言配置 作战能力 跨领域
下载PDF
基于模型驱动的密码算法可视化开发平台研究
11
作者 肖超恩 刘昌俊 +2 位作者 董秀则 王建新 张磊 《密码学报(中英文)》 CSCD 北大核心 2024年第2期357-370,共14页
针对密码算法开发平台普适性差、无法跨平台的问题,本文采用模型驱动实现密码算法开发的方法,设计了一种基于模型驱动的密码算法可视化开发平台,提出了一种基于模型驱动的密码算法开发的领域语言—MCL密码元语言;实现了基于模型的代码... 针对密码算法开发平台普适性差、无法跨平台的问题,本文采用模型驱动实现密码算法开发的方法,设计了一种基于模型驱动的密码算法可视化开发平台,提出了一种基于模型驱动的密码算法开发的领域语言—MCL密码元语言;实现了基于模型的代码生成器和代码映射器.实验证明,该开发平台仅需要开发者拖拽图形块的操作就可以实现密码算法模型的建立,然后平台可以根据建立的密码算法模型生成不同编程环境下的代码.平台实现了C和python的代码映射器模块,密码算法模型可快速映射为C、python代码.平台有较好的实用性,开发者的密码算法实现过程简洁、高效,不同编程环境下的代码均可以通过平台自动生成,提高了密码算法实现的跨平台性. 展开更多
关键词 密码算法实现 模型驱动 领域专用语言(DSL) 代码生成技术
下载PDF
基于语言模型的蛋白质结构域边界预测方法
12
作者 张贵军 汪乾梁 彭春祥 《浙江工业大学学报》 CAS 北大核心 2024年第5期521-529,共9页
蛋白质结构域对于蛋白质结构和功能研究具有重要意义。针对目前从头预测蛋白质结构域的方法普遍存在精度不高、耗费资源多等问题,提出了一种基于语言模型的蛋白质结构域边界预测方法DomTransformer,该方法基于蛋白质结构分类数据库(CATH... 蛋白质结构域对于蛋白质结构和功能研究具有重要意义。针对目前从头预测蛋白质结构域的方法普遍存在精度不高、耗费资源多等问题,提出了一种基于语言模型的蛋白质结构域边界预测方法DomTransformer,该方法基于蛋白质结构分类数据库(CATH)、蛋白质结构预测关键评估(CASP)竞赛数据,以及在AFDB(AlphaFold protein structure database)基础上建立的域数据库等共同构建数据集,搭建了基于Transformer网络架构和稀疏多头自注意力机制的网络模型,引入了新的特征、接触数和域级MSA(Domain multiple sequence alignment),通过直接预测结构域边界来解决数据不平衡等问题。在独立测试集上的测试结果表明了DomTransformer的有效性。 展开更多
关键词 蛋白质结构域 语言模型 从头预测
下载PDF
基于半监督学习的域适应实体解析算法
13
作者 戴超凡 丁华华 《计算机科学》 CSCD 北大核心 2024年第9期214-222,共9页
实体解析旨在查找两个数据实体是否引用同一实体,是许多自然语言处理任务中的一项基本任务。现有的基于深度学习的实体解析解决方案通常需要大量的标注数据,即使利用预训练的语言模型进行训练,仍然需要数千个标签才能达到令人满意的准... 实体解析旨在查找两个数据实体是否引用同一实体,是许多自然语言处理任务中的一项基本任务。现有的基于深度学习的实体解析解决方案通常需要大量的标注数据,即使利用预训练的语言模型进行训练,仍然需要数千个标签才能达到令人满意的准确性。现实场景中,这些标注数据并不容易获得。针对上述问题,提出了一个基于半监督学习的域适应实体解析模型。首先,在源域上训练一个分类器,然后利用域适应减小源域和目标域的分布差异,同时用数据增强后的目标域软伪标签加入源域迭代训练,从而实现从源域到目标域的知识迁移。在13个来自相同或不同领域的数据集上对所提模型进行了对比实验和消融实验,实验结果表明,与无监督基线模型相比,所提模型在多个数据集上的F1值平均提升了2.84%,9.16%和7.1%;与有监督基线模型相比,所提模型只需要20%~40%的标签就可以达到与有监督学习相当的性能。消融实验进一步证明了所提模型的有效性,其总体上可以获得更好的实体解析结果(相关代码已开源1))。 展开更多
关键词 实体解析 域适应 伪标签 预训练语言模型 数据增强
下载PDF
融合领域词典嵌入的航空不安全事件命名实体识别
14
作者 许雅玺 孟天宇 +1 位作者 王欣 刘炳南 《科学技术与工程》 北大核心 2024年第8期3284-3290,共7页
针对航空不安全事件领域命名实体识别任务,以航空安全信息周报为数据源,分析并构建航空不安全事件命名实体识别数据集和领域词典。为解决传统命名实体识别模型对于捕获领域实体边界性能较差的问题,基于BERT(bidirectional encoder repre... 针对航空不安全事件领域命名实体识别任务,以航空安全信息周报为数据源,分析并构建航空不安全事件命名实体识别数据集和领域词典。为解决传统命名实体识别模型对于捕获领域实体边界性能较差的问题,基于BERT(bidirectional encoder representations from transformers)预训练语言模型提出融合领域词典嵌入的领域语义信息增强的方法。在自建数据集上进行多次对比实验,结果表明:所提出的方法可以进一步提升实体边界的识别率,相较于传统的双向长短期记忆网络-条件随机场(bi-directional long short term memory-conditional random field,BiLSTM-CRF)命名实体识别模型,性能提升约5%。 展开更多
关键词 航空不安全事件 领域词典 命名实体识别 预训练语言模型
下载PDF
Modelica语言及其多领域统一建模与仿真机理 被引量:119
15
作者 赵建军 丁建完 +1 位作者 周凡利 陈立平 《系统仿真学报》 CAS CSCD 北大核心 2006年第z2期570-573,共4页
详细介绍了Modelica语言及其主要特点,系统地阐述了Modelica语言的多领域统一建模与仿真原理,分析了Modelica语言适合于复杂系统建模的内在原因,探讨了基于Modelica语言的复杂产品建模方法,综述了基于Modelica语言的建模仿真工具研究现... 详细介绍了Modelica语言及其主要特点,系统地阐述了Modelica语言的多领域统一建模与仿真原理,分析了Modelica语言适合于复杂系统建模的内在原因,探讨了基于Modelica语言的复杂产品建模方法,综述了基于Modelica语言的建模仿真工具研究现状,总结了采用Modelica语言进行多领域统一建模的优势。 展开更多
关键词 多领域 统一建模 协同仿真 Modlica
下载PDF
面向业务自动化转型的铁路运输调度业务数据模型构建方法研究
16
作者 郑然斐 孟令云 +3 位作者 苗建瑞 蒋熙 廖正文 潘钰雯 《铁道运输与经济》 北大核心 2024年第6期73-80,共8页
铁路运输生产调度业务自动化、智能化转型改进是推动铁路运输效率提升的重要手段。探讨业务转型过程中对业务各场景各颗粒度控制逻辑和信息流通的梳理建模方法,从而推动业务高效转型,在对运输生产调度业务深入分析的基础上,抽象了调度... 铁路运输生产调度业务自动化、智能化转型改进是推动铁路运输效率提升的重要手段。探讨业务转型过程中对业务各场景各颗粒度控制逻辑和信息流通的梳理建模方法,从而推动业务高效转型,在对运输生产调度业务深入分析的基础上,抽象了调度业务组织的关键影响要素和关键关联关系,梳理了工作流、信息流、信息流通机制与上述抽象的调度业务要素及这些要素间关联关系的对应关系,设计了一种刻画业务工作流、信息流和信息流通机制的数据模型构建方法。考虑到铁路运输生产调度业务分散协作、多级耦合的多层级多场景特点,设计了工作流、信息流数据模型的分化、细化关联机制,以实现对不同层级、场景调度业务的数据模型间有机关系的刻画,最后结合实际调度业务进行了应用分析。结果表明该方法构建的数据模型可以规范调度业务在不同场景、不同颗粒度下的控制逻辑和信息流通,从而为业务自动化转型过程中的信息系统开发和作业流程再造提供参考。 展开更多
关键词 调度指挥 数据模型 工作流 信息流 统一建模语言
下载PDF
基于知识图谱和预训练语言模型深度融合的可解释生物医学推理
17
作者 徐寅鑫 杨宗保 +2 位作者 林宇晨 胡金龙 董守斌 《北京大学学报(自然科学版)》 EI CAS CSCD 北大核心 2024年第1期62-70,共9页
基于预训练语言模型(LM)和知识图谱(KG)的联合推理在应用于生物医学领域时,因其专业术语表示方式多样、语义歧义以及知识图谱存在大量噪声等问题,联合推理模型并未取得较好的效果。基于此,提出一种面向生物医学领域的可解释推理方法DF-... 基于预训练语言模型(LM)和知识图谱(KG)的联合推理在应用于生物医学领域时,因其专业术语表示方式多样、语义歧义以及知识图谱存在大量噪声等问题,联合推理模型并未取得较好的效果。基于此,提出一种面向生物医学领域的可解释推理方法DF-GNN。该方法统一了文本和知识图谱的实体表示方式,利用大型生物医学知识库构造子图并进行去噪,改进文本和子图实体的信息交互方式,增加对应文本和子图节点的直接交互,使得两个模态的信息能够深度融合。同时,利用知识图谱的路径信息对模型推理过程提供了可解释性。在公开数据集MedQA-USMLE和MedMCQA上的测试结果表明,与现有的生物医学领域联合推理模型相比,DF-GNN可以更可靠地利用结构化知识进行推理并提供解释性。 展开更多
关键词 生物医学 预训练语言模型 知识图谱 联合推理
下载PDF
多领域建模语言Modelica类型解析研究与实现 被引量:5
18
作者 吴民峰 吴义忠 +1 位作者 周凡利 陈立平 《计算机工程与应用》 CSCD 北大核心 2006年第25期80-83,共4页
类型解析是编译器开发的一项重要工作,也是语义分析的一个最重要组成部分。基于Modelica建模语言,研究了编译器类型系统的作用域、类型检查等关键技术,提出了类型系统的解析和查找机制,实现了类型检查及错误处理机制,并在多领域物理系... 类型解析是编译器开发的一项重要工作,也是语义分析的一个最重要组成部分。基于Modelica建模语言,研究了编译器类型系统的作用域、类型检查等关键技术,提出了类型系统的解析和查找机制,实现了类型检查及错误处理机制,并在多领域物理系统建模与仿真平台MWorks系统中得到应用。 展开更多
关键词 类型系统 modelICA 多领域 建模语言
下载PDF
多体动力学模型的Modelica语言建模 被引量:3
19
作者 刘俊 黄运保 +1 位作者 陈立平 王启富 《中国机械工程》 EI CAS CSCD 北大核心 2010年第9期1088-1093,共6页
对Adams多体模型结构及Modelica模型的转换方法进行了研究。对多体动力学模型结构及建模方式进行分析,根据Adams多体模型结构设计了对应的Modelica多体模型结构。研究了Adams多体模型各组件包含的信息,以及与Modelica模型的异同,提出了... 对Adams多体模型结构及Modelica模型的转换方法进行了研究。对多体动力学模型结构及建模方式进行分析,根据Adams多体模型结构设计了对应的Modelica多体模型结构。研究了Adams多体模型各组件包含的信息,以及与Modelica模型的异同,提出了各多体组件的转换方法。最后给出了多体模型转换验证实例与结果。该研究有助于提高多领域仿真系统的多体建模效率及与传统多体系统的兼容性。 展开更多
关键词 多领域统一建模 modelICA语言 多体动力学模型 模型转换
下载PDF
问答式林业预训练语言模型ForestBERT
20
作者 谭晶维 张怀清 +2 位作者 刘洋 杨杰 郑东萍 《林业科学》 EI CAS CSCD 北大核心 2024年第9期99-110,共12页
【目的】针对林业文本利用率低、通用领域预训练语言模型对林业知识理解不足以及手动标注数据耗时费力等问题,基于大量林业文本,提出一种融合林业领域知识的预训练语言模型,并通过自动标注训练数据,高效实现林业抽取式问答,为林业决策... 【目的】针对林业文本利用率低、通用领域预训练语言模型对林业知识理解不足以及手动标注数据耗时费力等问题,基于大量林业文本,提出一种融合林业领域知识的预训练语言模型,并通过自动标注训练数据,高效实现林业抽取式问答,为林业决策管理提供智能化信息服务。【方法】首先,基于网络爬虫技术构建包含术语、法律法规和文献3个主题的林业语料库,使用该语料库对通用领域预训练语言模型BERT进行继续预训练,再通过掩码语言模型和下一句预测这2个任务进行自监督学习,使BERT能够有效地学习林业语义信息,得到具有林业文本通用特征的预训练语言模型ForestBERT。然后,对预训练语言模型mT5进行微调,实现样本的自动标注,通过人工校正后,构建包含3个主题共2280个样本的林业抽取式问答数据集。基于该数据集对BERT、RoBERTa、MacBERT、PERT、ELECTRA、LERT 6个通用领域的中文预训练语言模型以及本研究构建的ForestBERT进行训练和验证,以明确ForestBERT的优势。为探究不同主题对模型性能的影响,分别基于林业术语、林业法律法规、林业文献3个主题数据集对所有模型进行微调。将ForestBERT与BERT在林业文献中的问答结果进行可视化比较,以更直观展现ForestBERT的优势。【结果】ForestBERT在林业领域的抽取式问答任务中整体表现优于其他6个对比模型,与基础模型BERT相比,精确匹配(EM)分数和F1分数分别提升1.6%和1.72%,在另外5个模型的平均性能上也均提升0.96%。在各个模型最优划分比例下,ForestBERT在EM上分别优于BERT和其他5个模型2.12%和1.2%,在F1上分别优于1.88%和1.26%。此外,ForestBERT在3个林业主题上也均表现优异,术语、法律法规、文献任务的评估分数分别比其他6个模型平均提升3.06%、1.73%、2.76%。在所有模型中,术语任务表现最佳,F1的平均值达到87.63%,表现较差的法律法规也达到82.32%。在文献抽取式问答任务中,ForestBERT相比BERT可提供更准确、全面的答案。【结论】采用继续预训练的方式增强通用领域预训练语言模型的林业专业知识,可有效提升模型在林业抽取式问答任务中的表现,为林业文本和其他领域的文本处理和应用提供一种新思路。 展开更多
关键词 林业文本 BERT 预训练语言模型 特定领域预训练 抽取式问答任务 自然语言处理
下载PDF
上一页 1 2 54 下一页 到第
使用帮助 返回顶部