期刊文献+
共找到712篇文章
< 1 2 36 >
每页显示 20 50 100
Training and Implementation of Subjective Questions Scoring System Based on the Baidu Qianfan Model Platform
1
作者 Xiaoyun Zhu 《Journal of Contemporary Educational Research》 2024年第11期227-232,共6页
Leveraging the Baidu Qianfan model platform,this paper designs and implements a highly efficient and accurate scoring system for subjective questions,focusing primarily on questions in the field of computer network te... Leveraging the Baidu Qianfan model platform,this paper designs and implements a highly efficient and accurate scoring system for subjective questions,focusing primarily on questions in the field of computer network technology.The system enhances the foundational model by utilizing Qianfan’s training tools and integrating advanced techniques,such as supervised fine-tuning.In the data preparation phase,a comprehensive collection of subjective data related to computer network technology is gathered,cleaned,and labeled.During model training and evaluation,optimal hyperparameters and tuning strategies are applied,resulting in a model capable of scoring with high accuracy.Evaluation results demonstrate that the proposed model performs well across multiple dimensions-content,expression,and development scores-yielding results comparable to those of manual scoring. 展开更多
关键词 Subjective score Natural language processing Deep learning Baidu Qianfan large model platform Supervised fine-tuning Model training and evaluation
在线阅读 下载PDF
Methodology of the Comprehensive Teacher Training
2
作者 Dzhumanova Lola 《Sino-US English Teaching》 2017年第7期421-424,共4页
Pedagogical tradition regarding musical education in Russia is one of the country culture achievements.Arrangement of quality musical pedagogy has been a matter of special importance within half a century history ofmu... Pedagogical tradition regarding musical education in Russia is one of the country culture achievements.Arrangement of quality musical pedagogy has been a matter of special importance within half a century history ofmusic education system forming. Training of music teachers is particularly important in the field of music theory,as the comprehensive musician training is specific for Russian music education, where besides playing a musicinstrument he must possess knowledge from the music theory field, make sense of music pieces, and understandand analyze them. The conservatism in teaching traditions preservation began interfering with changing priorities ofeducation. As a result, stagnation in the field of musical pedagogy was outlined in the early XXI century. This hasaffected teaching music and theoretical subjects, which became heavy, boring, and unpopular due to conservatismin teaching methods and focus on students' motivation. As professional regards, there has been an imbalance in theability to play the instrument and the inability to navigate in the music literature. Implementation of innovativeapproaches to teaching issues and appropriate teachers training has become a necessary solution to the problem.The report considers innovations in teachers training traditions on example of the leading musical institutionMoscow Tchaikovsky Conservatory. 展开更多
关键词 PEDAGOGICAL training of musicologists COMPREHENSIVE approach innovative FORMS of training multidisciplinary nature
在线阅读 下载PDF
EZH2 identifies the precursors of human natural killer cells with trained immunity
3
作者 Chen Zhang Jie Yin +14 位作者 Jian Zheng Jun Xiao Jiajian Hu Yudong Su Kaichen Zhou Yingchi Zhang Xuzhen Zhang Hong Zhang Qian Sun Yang Wang Wenwen Yu Feng Wei Qiang Zhao Long Li Xiubao Ren 《Cancer Biology & Medicine》 SCIE CAS CSCD 2021年第4期1021-1039,共19页
Objective:Trained immunity of natural killer(NK)cells has shown great potential in the treatment of cancers by eliciting enhanced effector responses to restimulation by cytokines or cancer cells for long time periods ... Objective:Trained immunity of natural killer(NK)cells has shown great potential in the treatment of cancers by eliciting enhanced effector responses to restimulation by cytokines or cancer cells for long time periods after preactivation.However,the human NK cells responsible for the generation and maintenance of trained immunity are largely unknown.We hypothesized that heterogeneous human NK cells would respond differentially to stimulation with a combination of IL-12,IL-15,and IL-18,and that an NK cell subset might exist that is mainly responsible for the induction of trained immunity.On the basis of our hypothesis,we aimed to identify the subset from which cytokine-trained human NK cells originate and to explore possible regulatory targets for drug intervention.Methods:Flow cytometry assays were performed to analyze the functions of cytokine-trained NK cells and examine cell division and protein expression in NK cell subsets.Single-cell RNA sequencing(scRNA-seq)plus TotalSeq™technology was used to track the heterogeneity of NK cells during the induction of trained immunity.Results:Traditional developmental markers for peripheral NK cells were unable to identify the precursors of human NK cells with trained immunity.Therefore,we used scRNA-seq plus TotalSeq™technology to track the heterogeneity of NK cells during the induction of trained immunity and identified a unique cluster of CD57−NKG2A+EZH2+IFNG+MKI67+IL12R+IL15R+IL18R+NK cells.Enrichment and pseudotime trajectory analyses suggested that this cluster of NK cells contained the precursor of trained NK cells.We then used flow cytometry to further investigate the role of EZH2 in trained NK precursors and found that CD57−NKG2A+EZH2+NK cells had faster cell cycles and an enhanced trained phenotype,and EZH2 inhibition significantly impaired the induction of trained immunity in NK cells.These results suggested that EZH2 is a unique epigenetic marker of precursors of human NK cells with trained immunity.Conclusions:Our work revealed human NK heterogeneity in the induction of trained immunity,identified the precursor subset for trained NK cells,and demonstrated the critical role of EZH2 in the induction of trained immunity in human NK cells. 展开更多
关键词 Natural killer cells trained immunity PRECURSOR EZH2 cell cycle
在线阅读 下载PDF
Analyzing the Impact of Different Factors on Software Cost Estimation in Today’s Scenario
4
作者 Deepa Gangwani Saurabh Mukherjee 《Journal of Software Engineering and Applications》 2015年第5期245-251,共7页
Software cost estimation is a main concern of the software industry. However, the fact is also that in today’s scenario, software industries are more interested in other issues like new technologies in the market, sh... Software cost estimation is a main concern of the software industry. However, the fact is also that in today’s scenario, software industries are more interested in other issues like new technologies in the market, shorter development time, skill shortage etc. They are actually deviating from critical issues to routine issues. Today, people expect high quality products at very low costs and same is the goal of software engineering. An accuracy in software cost estimation has a direct impact on company’s reputation and also affects the software investment decisions. Accurate cost estimation can minimize the unnecessary costs and increase the productivity and efficiency of the company. The objective of this paper is to identify the existing methods of software cost estimation prevailing in the market and analyzing some of the important factors impacting the software cost estimation process. In order to achieve the objective, a survey was conducted to find out: ● Nature of projects that companies prefer. ● Impact of training on employees in software cost estimation. ● How many people review the estimated cost? ● How much risk buffer the company keeps for future prospects? 展开更多
关键词 SofTWARE COST Estimation training Research method nature of Project SofTWARE Review
在线阅读 下载PDF
Continuity of Natural Science Education in the System of“School-University”
5
作者 Shuinshina S.M Alpeisov E.A. +1 位作者 Zhakupov A.A. Burunbetova K.K. 《Sino-US English Teaching》 2018年第5期247-252,共6页
At the present time,new state compulsory standards for school education,standard curricula,and curricula have been developed.The programs of higher and postgraduate education do not take into account changes in educat... At the present time,new state compulsory standards for school education,standard curricula,and curricula have been developed.The programs of higher and postgraduate education do not take into account changes in educational curricula of secondary education.The future specialists of pedagogical universities are teachers of the school.Therefore,graduates of universities on pedagogical specialties,including scientific direction,should be profoundly aware of the content of subjects studied in secondary education.It is very important to understand the need to ensure the continuity of educational curricula of higher pedagogical and general secondary education in conditions of updating the content of education.The relevance of the research topic is due to the lack of continuity in the educational programs of higher pedagogical and general secondary education in conditions of updating the content of education;discrepancy of professional competences of young pedagogical staff,graduates of higher educational institutions,to the conditions of the updated content of school education;the lack of fundamental and applied research to date,considering the theoretical and methodological foundations for ensuring the continuity of natural science education in the“school-university”system in the context of updating the content of education;the need for the research in the context of the implementation of the State Program for the Development of Education of the Republic of Kazakhstan for 2016-2019,which indicates an increase in the prestige of the profession of teachers and an increase in their quality.These theoretical and methodological bases for ensuring the continuity of educational programs of higher pedagogical and secondary education in conditions of updating the content of education will make it possible to introduce changes in the educational curricula of higher and postgraduate education for the training of pedagogical personnel. 展开更多
关键词 CONTINUITY in the system“school-university” natural SCIENCES and mathematics ensuring the CONTINUITY of educational content training of TEACHERS state OBLIGATORY standard of EDUCATION curriculum
在线阅读 下载PDF
Research on Subject Nature and Objective of Physical Education Training
6
作者 Pengfei Wen Menghua Wang 《Journal of Contemporary Educational Research》 2019年第6期31-34,共4页
Physical education and training is considered as a new subject among the same type of discipline.It has a short history background and is relatively young.However,the influence of the discipline of physical education ... Physical education and training is considered as a new subject among the same type of discipline.It has a short history background and is relatively young.However,the influence of the discipline of physical education and training must not be ignored by simply blaming it as the youngest discipline.In addition,especially in the case of rapid economic development and rapid improvement in living standards today,the influence of this discipline is relatively strong.Therefore,this thesis mainly studies the subject nature and object of physical education training.The research object of this subject can be studied from three key points,such as:the curriculum of the subject,the research method and the research of the subject of education.The comprehensive development of physical education and physical education is a comprehensive discipline. 展开更多
关键词 PHYSICAL EDUCATION PHYSICAL training SUBJECT nature
在线阅读 下载PDF
The Joint Model of Multi-Intent Detection and Slot Filling Based on Bidirectional Interaction Structure
7
作者 WANG Changjing ZENG Xianghui +2 位作者 WANG Yuxin SUN Yuxin ZUO Zhengkang 《Wuhan University Journal of Natural Sciences》 2025年第1期21-31,共11页
Intent detection and slot filling are two important components of natural language understanding.Because their relevance,joint training is often performed to improve performance.Existing studies mostly use a joint mod... Intent detection and slot filling are two important components of natural language understanding.Because their relevance,joint training is often performed to improve performance.Existing studies mostly use a joint model of multi-intent detection and slot-filling with unidirectional interaction,which improves the overall performance of the model by fusing the intent information in the slot-filling part.On this basis,in order to further improve the overall performance of the model by exploiting the correlation between the two,this paper proposes a joint multi-intent detection and slot-filling model based on a bidirectional interaction structure,which fuses the intent encoding information in the encoding part of slot filling and fuses the slot decoding information in the decoding part of intent detection.Experimental results on two public multi-intent joint training datasets,MixATIS and MixSNIPS,show that the bidirectional interaction structure proposed in this paper can effectively improve the performance of the joint model.In addition,in order to verify the generalization of the bidirectional interaction structure between intent and slot,a joint model for single-intent scenarios is proposed on the basis of the model in this paper.This model also achieves excellent performance on two public single-intent joint training datasets,CAIS and SNIPS. 展开更多
关键词 natural language understanding multi-intent detection slot filling bidirectional interaction joint training
原文传递
Extraction of Robot Primitive Control Rules from Natural Language Instructions 被引量:1
8
作者 Guang-Hong Wang Ping Jiang Zu-Ren Feng 《International Journal of Automation and computing》 EI 2006年第3期282-290,共9页
A support vector rule based method is investigated for the construction of motion controllers via natural language training. It is a two-phase process including motion control information collection from natural langu... A support vector rule based method is investigated for the construction of motion controllers via natural language training. It is a two-phase process including motion control information collection from natural language instructions, and motion information condensation with the aid of support vector machine (SVM) theory. Self-organizing fuzzy neural networks are utilized for the collection of control rules, from which support vector rules are extracted to form a final controller to achieve any given control accuracy. In this way, the number of control rules is reduced, and the structure of the controller tidied, making a controller constructed using natural language training more appropriate in practice, and providing a fundamental rule base for high-level robot behavior control. Simulations and experiments on a wheeled robot are carried out to illustrate the effectiveness of the method. 展开更多
关键词 Support vector machines (SVMs) fuzzy neural networks motion primitives motion controller language instruction based training natural language programming.
在线阅读 下载PDF
Dynamic characteristics and sensitivities analysis of a power turret gear train
9
作者 Xu Lijiao Chen Nan 《仪器仪表学报》 EI CAS CSCD 北大核心 2014年第S1期174-179,共6页
The dynamic transmission characteristics and the sensitivities of the three stage idler gear system of the new NC power turret are studied in the paper. Considering the strongly nonlinear factors such as the periodica... The dynamic transmission characteristics and the sensitivities of the three stage idler gear system of the new NC power turret are studied in the paper. Considering the strongly nonlinear factors such as the periodically time-varying mesh stiffness, the nonlinear tooth backlash, the lump-parameter model of the gear system is developed with one rotational and two translational freedoms of each gear. The eigen-values and eigenvectors are derived and analyzed on the basis of the real modal theory. The sensitivities of natural frequencies to design parameters including supporting and meshing stiffnesses, gear masses, and moments of inertia by the direct differential method are also calculated. The results show the quantitative and qualitative impact of the parameters to the natural characteristics of the gear system. Furthermore, the periodic steady state solutions are obtained by the numerical approach based on the nonlinear model. These results are employed to gain insights into the primary controlling parameters, to forecast the severity of the dynamic response, and to assess the acceptability of the gear design. 展开更多
关键词 GEAR TRAIN dynamic response NATURAL PROPERTY sensitivity
在线阅读 下载PDF
JOURNAL OF CHENGDU UNIVERSITY Social Sciences No. 3 1990 Sichuan Association of Editors for College and University Journals
10
《成都大学学报(社会科学版)》 1990年第3期97-96,共2页
关键词 college TEACHER Style training Distance Behaviour nature LEADERSHIP regarding
在线阅读 下载PDF
论党内法规编纂的性质和条件——兼论党内法规法典化编纂的基本思路
11
作者 秦前红 刘韵笛 《西北师大学报(社会科学版)》 北大核心 2025年第2期14-23,共10页
党内法规编纂有别于党内法规法典化,是指对彼此间呈高关联性、整体上可再体系化的若干现行党内法规加以研究,根据集成原则和现实需要施以增删纂修,以消解既存党内法规规范中的阙漏和冲突,最终按照一定逻辑理路和体例结构编制一部较前更... 党内法规编纂有别于党内法规法典化,是指对彼此间呈高关联性、整体上可再体系化的若干现行党内法规加以研究,根据集成原则和现实需要施以增删纂修,以消解既存党内法规规范中的阙漏和冲突,最终按照一定逻辑理路和体例结构编制一部较前更为系统化的新党内法规的活动。在性质和功能上,党内法规编纂是党内法规制定的下位概念,是党内法规汇编的实质升级,是党内法规法典化的重要但非唯一途径。启动和实施党内法规编纂应有计划、按步骤进行,须满足认知、制度、技术等主客观条件。直面“如何推进党内法规法典化编纂”的时代课题,须遵循党内法规制度建设一般规律,树立科学的党内法规“法典观”,把准党内法规法典化发展进程,以居中式法典化为基础模式、以党内法规清理为前端环节、以党的组织法规为探索对象、以民法典编纂为经验参考,稳步开展党内法规法典化编纂预备工作。 展开更多
关键词 党内法规 编纂 法典化 再体系化 性质
在线阅读 下载PDF
新时代地方高校研究生培养特区建设的举措与实践——以石油与天然气工程类专业为例
12
作者 王志华 杨二龙 +3 位作者 孙巍 王凤娇 张军 刘音颂 《高教学刊》 2025年第7期27-30,共4页
以石油与天然气行业高层次人才培养目标为导向,分别从思政育人、研究生培养方向、理论教学体系、专业实践教学、导师人才队伍以及质量监控评价体系出发,探讨新时代地方高校研究生培养特区建设的路径,明确石油天然气产业全链条高层次人... 以石油与天然气行业高层次人才培养目标为导向,分别从思政育人、研究生培养方向、理论教学体系、专业实践教学、导师人才队伍以及质量监控评价体系出发,探讨新时代地方高校研究生培养特区建设的路径,明确石油天然气产业全链条高层次人才培养方向,建立新时代石油与天然气行业研究生教育理论体系、实践体系与质控标准。建设经验和做法助力东北石油大学研究生培养水平的提升,同时具有在同类高校推广应用的前景。 展开更多
关键词 研究生培养特区 研究生教育 优势学科 新时代 石油与天然气工程
在线阅读 下载PDF
A NUMERICAL STUDY OF THE TRAIN-INDUCED UNSTEADY AIRFLOW IN A SUBWAY TUNNEL WITH NATURAL VENTILATION DUCTS USING THE DYNAMIC LAYERING METHOD 被引量:17
13
作者 HUANG Yuan-dong GAO Wei KIM Chang-Nyung 《Journal of Hydrodynamics》 SCIE EI CSCD 2010年第2期164-172,共9页
The objective of this study is to investigate numerically the characteristics of train-induced unsteady airflow in a subway tunnel with natural ventilation ducts.A three-dimensional numerical model using the dynamic l... The objective of this study is to investigate numerically the characteristics of train-induced unsteady airflow in a subway tunnel with natural ventilation ducts.A three-dimensional numerical model using the dynamic layering method for the moving boundary of a train is first developed,and then it is validated against the model tunnel experimental data.With the tunnel and subway train geometries in the numerical model exactly the same as those in the model tunnel experimental test,but with the ventilation ducts being connected to the tunnel ceiling and a barrier placed at the tunnel outlet,the three-dimensional train-induced unsteady tunnel flows are numerically simulated.The computed distributions of the pressure and the air velocity in the tunnel as well as the time series of the mass flow rate at the ventilation ducts reveal the impact of the train motion on the exhaust and suction of the air through ventilation ducts and the effects of a barrier placed at the tunnel outlet on the duct ventilation performance.As the train approaches a ventilation duct,the air is pushed out of the tunnel through the duct.As the train passes the ventilation duct,the exhaust flow in the duct is changed rapidly to the suction flow.After the train passes the duct,the suction mass flow rate at the duct decreases with time since the air pressure at the opening of the duct is gradually recovered with time.A drastic change in the mass flow rate at a ventilation duct while a train passes the corresponding ventilation duct,causes a change in the exhaust mass flow rate at other ventilation ducts.Also,when a barrier is placed at the tunnel outlet,the air volume discharge rate at each ventilation duct is greatly increased,i.e.,the barrier placed at the tunnel outlet can improve remarkably the ventilation performance through each duct. 展开更多
关键词 train-induced flow subway tunnel natural ventilation duct ventilation performance dynamic layering method
原文传递
A NUMERICAL ANALYSIS OF THE VENTILATION PERFORMANCE FOR DIFFERENT VENTILATION STRATEGIES IN A SUBWAY TUNNEL 被引量:8
14
作者 HUANG Yuan-dong LI Chan KIM Chang Nyung 《Journal of Hydrodynamics》 SCIE EI CSCD 2012年第2期193-201,共9页
An unsteady three-dimensional analysis of the ventilation performance is carried out for different ventilation strategies to find out a ventilation method with a high performance in a subway tunnel.The natural ventila... An unsteady three-dimensional analysis of the ventilation performance is carried out for different ventilation strategies to find out a ventilation method with a high performance in a subway tunnel.The natural ventilation performance associated with a train-induced air flow in a subway tunnel is examined.The dynamic layering method is used to consider the moving boundary of a train in the current CFD method.The geometries of the modeled tunnel and the subway train are partially based on those of the Seoul subway.The effects of the structure of the ventilation duct and the geometry of the partitions on the ventilation performance are evaluated.The results show that the combined ventilation ducts (to be designed),and the partitioning blocks installed along the middle of tunnel (already in existences) are helpful for air exchange.This study can provide some guidance for the design of ventilation ducts in a subway system. 展开更多
关键词 ventilation performance subway tunnel train-induced air flow natural ventilation duct numerical analysis
原文传递
基于预训练模型与双向注意力流的抽取式阅读理解模型
15
作者 文勇军 吴金铭 梅硕 《首都师范大学学报(自然科学版)》 2025年第2期1-11,共11页
针对服务机器人的抽取式阅读理解任务中出现答案抽取准确度不高的问题,构建了基于预训练模型与双向注意力流的抽取式阅读理解模型。该模型首先采用预训练模型来提取问题与文档上下文的浅层联合语义表征;其次利用双向注意力网络来加强特... 针对服务机器人的抽取式阅读理解任务中出现答案抽取准确度不高的问题,构建了基于预训练模型与双向注意力流的抽取式阅读理解模型。该模型首先采用预训练模型来提取问题与文档上下文的浅层联合语义表征;其次利用双向注意力网络来加强特征交互和信息融合,得到问题与文档上下文的深层联合语义表征;最后结合浅层和深层的联合语义表征,通过排序、滤错和定位操作完成对答案的抽取。在抽取式问答任务的斯坦福英文机器阅读理解数据集SQuAD 1.1和“讯飞杯”中文机器阅读理解数据集CMRC 2018上进行了实验。结果表明:与英文预训练语言模型BERT相比,该模型的性能指标EM和F1值分别提高了1.172%和1.194%;与中文预训练语言模型RoBERTa-wwm-ext相比,该模型的EM和F1值分别提高了1.336%和0.921%。 展开更多
关键词 自然语言处理 机器阅读理解 预训练模型 双向注意力流(BERT) RoBERTa-wwm-ext 答案抽取
在线阅读 下载PDF
视觉富文档理解预训练综述 被引量:1
16
作者 张剑 李晖 +2 位作者 张晟铭 吴杰 彭滢 《计算机科学》 北大核心 2025年第1期259-276,共18页
视觉富文档指语义结构不仅由文本内容决定,还与排版格式和表格结构等视觉元素相关的文档。现实生活中的票据理解和证件识别等应用场景,都需要对视觉富文档进行自动化的阅读、分析和处理。这一过程即为视觉富文档理解,属于自然语言处理... 视觉富文档指语义结构不仅由文本内容决定,还与排版格式和表格结构等视觉元素相关的文档。现实生活中的票据理解和证件识别等应用场景,都需要对视觉富文档进行自动化的阅读、分析和处理。这一过程即为视觉富文档理解,属于自然语言处理和计算机视觉的交叉领域。近年来,视觉富文档理解领域的预训练技术在打破下游任务的训练壁垒和提升模型表现上取得了重大的进展。然而,目前对现有的预训练模型的归纳总结和深入分析仍然有所欠缺。为此,对视觉富文档理解领域预训练技术的相关研究进行了全面总结。首先,介绍了预训练技术的数据预处理阶段,包括预训练数据集和光学字符识别引擎。然后,对预训练技术的模型预训练阶段进行了阐述,提炼出单模态表示学习、多模态特征融合和预训练任务3个关键的技术模块,并基于上述模块归纳了预训练模型之间的共性和差异。此外,简要介绍了多模态大模型在视觉富文档理解领域的应用。接着,对预训练模型在下游任务上的表现进行了对比分析。最后,探讨了预训练技术面临的挑战和未来的研究方向。 展开更多
关键词 文档智能 预训练模型 自然语言处理 计算机视觉 深度学习
在线阅读 下载PDF
影响预训练语言模型数据泄露的因素研究
17
作者 钱汉伟 彭季天 +4 位作者 袁明 高光亮 刘晓迁 王群 朱景羽 《信息安全研究》 北大核心 2025年第2期181-188,共8页
当前广泛使用的预训练语言模型是从海量训练语料中学习通用的语言表示.自然语言处理领域的下游任务在使用预训练语言模型后性能得到显著提升,但是深度神经网络过拟合现象使得预训练语言模型可能存在泄露训练语料隐私的风险.选用T5,GPT-2... 当前广泛使用的预训练语言模型是从海量训练语料中学习通用的语言表示.自然语言处理领域的下游任务在使用预训练语言模型后性能得到显著提升,但是深度神经网络过拟合现象使得预训练语言模型可能存在泄露训练语料隐私的风险.选用T5,GPT-2,OPT等广泛使用的预训练语言模型作为研究对象,利用模型反演攻击探索影响预训练语言模型数据泄露的因素.实验过程中利用预训练语言模型生成大量样本,以困惑度等指标选取最有可能发生数据泄露风险的样本进行验证,证明了T5等不同模型均存在不同程度的数据泄露问题;同一种模型,模型规模越大数据泄露可能性越大;添加特定前缀更容易获取泄露数据等问题.对未来数据泄露问题及其防御方法进行了展望. 展开更多
关键词 自然语言处理 预训练语言模型 隐私数据泄露 模型反演攻击 模型架构
在线阅读 下载PDF
基于预训练模型标记器重构的藏文分词系统
18
作者 杨杰 尼玛扎西 +2 位作者 仁青东主 祁晋东 才让东知 《计算机应用》 北大核心 2025年第4期1199-1204,共6页
针对现有的预训练模型在藏文分词任务中表现不佳的问题,提出一种建立重构标记器规范约束文本,随后重构藏文预训练模型的标记器以进行藏文分词任务的方法。首先,对原始文本进行规范化操作,以解决因语言混用等导致的错误切分的问题;其次,... 针对现有的预训练模型在藏文分词任务中表现不佳的问题,提出一种建立重构标记器规范约束文本,随后重构藏文预训练模型的标记器以进行藏文分词任务的方法。首先,对原始文本进行规范化操作,以解决因语言混用等导致的错误切分的问题;其次,对预训练模型进行音节粒度的标记器重构,使得切分单元与标注单元平行;最后,在利用改进的滑动窗口还原法完成黏着切分后,利用“词首、词中、词尾、孤立”(BMES)四元标注法建立Re-TiBERTBiLSTM-CRF模型,从而得到藏文分词系统。实验结果表明,重构标记器后的预训练模型在分词任务中明显优于原始预训练模型,而得到的系统拥有较高的藏文分词精确率,F1值最高可达97.15%,能够较好地完成藏文分词任务。 展开更多
关键词 藏语信息处理 藏文分词模型 预训练模型 自然语言处理 标记器重构
在线阅读 下载PDF
融合ERNIE的自然灾害舆情事理图谱构建及次生衍生事件探测研究
19
作者 李诗轩 王璐 +2 位作者 沈愿 陈烨 周禾深 《情报杂志》 北大核心 2025年第3期128-138,共11页
[研究目的]基于规则模版和大语言模型构建面向自然灾害的网络舆情事理图谱,结合情感分析进行舆情演化分析,进而实现次生衍生事件探测,为突发事件舆情管理与预警提供理论和实践参考。[研究方法]分别利用规则模板和预训练模型ERNIE提取关... [研究目的]基于规则模版和大语言模型构建面向自然灾害的网络舆情事理图谱,结合情感分析进行舆情演化分析,进而实现次生衍生事件探测,为突发事件舆情管理与预警提供理论和实践参考。[研究方法]分别利用规则模板和预训练模型ERNIE提取关系事件对,并利用K-means和BERTopic进行事件泛化,构建自然灾害网络舆情抽象事理图谱。同时利用情感词典进行情感演化分析,结合同类灾害事件的相似度计算,构建同类事件的网络舆情预测事理图谱。[研究结果/结论]自然灾害舆情抽象事理图谱能够分析事件的因果与顺承关系,揭示舆情演化过程中的关键事件。预测事理图谱可以辅助次生衍生事件探测,实现舆情预警。对比发现基于预训练模型ERNIE结合BERTopic构建的图谱抽象事件较少,事件集中于中心节点,而规则模板结合K-means得到的图谱抽象事件较多,关系较为分散。 展开更多
关键词 网络舆情 自然灾害 事理图谱 预训练模型ERNIE 情感演化分析 次生衍生事件探测
在线阅读 下载PDF
面向研究生招生咨询的中文Text-to-SQL模型
20
作者 王庆丰 李旭 +1 位作者 姚春龙 程腾腾 《计算机工程》 北大核心 2025年第3期362-368,共7页
研究生招生咨询是一种具有代表性的短时间高频次问答应用场景。针对现有基于词向量等方法的招生问答系统返回答案不够精确,以及每年需要更新问题库的问题,引入了基于文本转结构化查询语言(Text-to-SQL)技术的RESDSQL模型,可将自然语言... 研究生招生咨询是一种具有代表性的短时间高频次问答应用场景。针对现有基于词向量等方法的招生问答系统返回答案不够精确,以及每年需要更新问题库的问题,引入了基于文本转结构化查询语言(Text-to-SQL)技术的RESDSQL模型,可将自然语言问题转化为SQL语句后到结构化数据库中查询答案并返回。搜集了研究生招生场景中的高频咨询问题,根据3所高校真实招生数据,构建问题与SQL语句模板,通过填充模板的方式构建数据集,共有训练集1501条、测试集386条。将RESDSQL的RoBERTa模型替换为具有更强多语言生成能力的XLM-RoBERTa模型、T5模型替换为mT5模型,并在目标领域数据集上进行微调,在招生领域问题上取得了较高的准确率,在mT5-large模型上执行正确率为0.95,精确匹配率为1。与基于ChatGPT3.5模型、使用零样本提示的C3SQL方法对比,该模型性能与成本均更优。 展开更多
关键词 中文文本转结构化查询语言 自然语言查询 中文SQL语句生成 预训练模型 Text-to-SQL数据集
在线阅读 下载PDF
上一页 1 2 36 下一页 到第
使用帮助 返回顶部