[1]刘佳旭,班子恒,张艳菊.混合位置编码的议论文语篇要素识别模型[J].计算机技术与发展,2025,(10):139-147.[doi:10.20165/j.cnki.ISSN1673-629X.2025.0110]
　LIU Jia-xu,BAN Zi-heng,ZHANG Yan-ju.Hybrid Position Encoding Model for Discourse Element Identification in Argumentative Essays[J].,2025,(10):139-147.[doi:10.20165/j.cnki.ISSN1673-629X.2025.0110]
点击复制

混合位置编码的议论文语篇要素识别模型

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2025年10期
页码:
139-147
栏目:
人工智能
出版日期:
2025-10-10

文章信息/Info

Title:
Hybrid Position Encoding Model for Discourse Element Identification in Argumentative Essays
文章编号:
1673-629X(2025)10-0139-09
作者:
刘佳旭1; 班子恒1; 张艳菊2
1. 辽宁工程技术大学 软件学院,辽宁葫芦岛 125105; 2. 辽宁工程技术大学工商管理学院,辽宁葫芦岛 125105
Author(s):
LIU Jia-xu1; BAN Zi-heng1; ZHANG Yan-ju2
1. School of Software,Liaoning Technical University,Huludao 125105,China; 2. School of Business Administration,Liaoning Technical University,Huludao 125105,China
关键词:
语篇要素识别; 双向长短期记忆网络; 混合位置编码; 深度学习; 注意力机制
Keywords:
discourse element identification;  bidirectional long short- term memory;  hybrid position encoding;  deep learning;  attention mechanism
分类号:
TP391
DOI:
10.20165/j.cnki.ISSN1673-629X.2025.0110
摘要:
语篇要素识别在自动作文评分中发挥着重要作用,提高语篇要素识别的准确率有助于增强自动作文评分的效果以及可解释性。然而,语篇要素识别任务面临着上下文依赖和句子歧义性等挑战。传统的基于规则和特征工程的方法难以捕捉文本中复杂的语义信息和长距离依赖关系,而深度学习方法虽然能够自动学习文本特征,但仍然存在对关键位置信息利用不足的问题。针对上述问题,提出了一种混合位置编码的语篇要素识别模型,即HPE-BiLSTM(Hybrid Position Encoding Bidirectional Long Short-Term Memory)。该模型首先基于预训练的词向量获取句子表示,然后通过双向长短期记忆网络提取句子级特征。在句子级特征的基础上,采用混合的位置编码方案以确保关键位置信息的有效传递。最后,使用线性层和激活函数实现语篇要素识别。该模型在议论文数据集进行实验,并与Feature-based、BERT、BiLSTM、DiSA和DCRGNN五个模型进行比较。实验结果表明,HPE-BiLSTM模型的准确率达到了0.693,在语篇要素识别方面的F?分数为0.684,优于其他模型。
Abstract:
Discourse element identification plays an important role in automated essay scoring. Improving the accuracy of discourse element identification is helpful to improve the validity and explainability of automated essay scoring. However,discourse element identi-fication tasks face challenges such as context dependence and sentence ambiguity. Traditional rule-based and feature engineering methods are difficult to capture complex semantic information and long- distance dependencies in texts, while deep learning methods can automatically learn text features,but still have the problem of underutilization of key location information. To address these issues,HPE-BiLSTM (Hybrid Position Encoding Bidirectional Long Short- term Memory) is proposed. This model first obtains sentence representations using pre-trained word embeddings,and then extracts sentence-level features through a bidirectional long short-term memory network. Based on the sentence-level features,a hybrid position encoding scheme is applied to ensure the effective transmission of critical positional information. Finally,linear layers and activation functions are used to perform discourse element identification. The model is tested on an argumentative essay dataset and compared with five other models:Feature-based,BERT,BiLSTM,DiSA,and DCRGNN. Experimental results show that the HPE-BiLSTM model achieves an accuracy of 0.693 and an F? score of 0.684 in discourse element identification,outperforming the other models.

相似文献/References:

[1]周 康,曲卫东,杨艺琛.基于增强 BiLSTM 的网络文章核心实体识别[J].计算机技术与发展,2021,31(01):7.[doi:10. 3969 / j. issn. 1673-629X. 2021. 01. 002]
　ZHOU Kang,QU Wei-dong,YANG Yi-chen.Core Entity Recognition of Web Articles Based on Enhanced BiLSTM[J].,2021,31(10):7.[doi:10. 3969 / j. issn. 1673-629X. 2021. 01. 002]
[2]何阳宇,易晓宇,唐　亮,等.基于ＢＬＳＴＭ－ＡＴＴ的老挝语军事领域实体关系抽取[J].计算机技术与发展,2021,31(05):31.[doi:10. 3969 / j. issn. 1673-629X. 2021. 05. 006]
　,,et al.ＬａｏＥｎｔｉｔｙＲｅｌａｔｉｏｎＥｘｔｒａｃｔｉｏｎｉｎＭｉｌｉｔａｒｙＤｏｍａｉｎＢａｓｅｄｏｎＢＬＳＴＭ ａｎｄＡｔｔｅｎｔｉｏｎＭｅｃｈａｎｉｓｍ[J].,2021,31(10):31.[doi:10. 3969 / j. issn. 1673-629X. 2021. 05. 006]
[3]王卫红,吕红燕,曹玉辉,等.基于 BERT 的混合神经网络实体识别方法[J].计算机技术与发展,2021,31(08):100.[doi:10. 3969 / j. issn. 1673-629X. 2021. 08. 017]
　WANG Wei-hong,LYU Hong-yan,CAO Yu-hui,et al.A Hybrid Neural Network Entity Recognition Method Based on BERT Model[J].,2021,31(10):100.[doi:10. 3969 / j. issn. 1673-629X. 2021. 08. 017]
[4]孙安亮,时宏伟,王金策.基于字符与单词嵌入的航空安全命名实体识别[J].计算机技术与发展,2022,32(09):148.[doi:10. 3969 / j. issn. 1673-629X. 2022. 09. 023]
　SUN An-liang,SHI Hong-wei,WANG Jin-ce.Named Entity Recognition Based on Character and Word Embedding in Aviation Safety[J].,2022,32(10):148.[doi:10. 3969 / j. issn. 1673-629X. 2022. 09. 023]
[5]胡慧婷,李建平,董振荣,等.基于 BERT 模型的教育技术学领域实体抽取[J].计算机技术与发展,2022,32(10):164.[doi:10. 3969 / j. issn. 1673-629X. 2022. 10. 027]
　HU Hui-ting,LI Jian-ping,DONG Zhen-rong,et al.Named Entity Recognition Method in Educational Technology Field Based on BERT[J].,2022,32(10):164.[doi:10. 3969 / j. issn. 1673-629X. 2022. 10. 027]
[6]李春生,田梦晴,张可佳.基于 Bi-LSTM 网络的管道异常数据检测方法[J].计算机技术与发展,2023,33(06):215.[doi:10. 3969 / j. issn. 1673-629X. 2023. 06. 032]
　LI Chun-sheng,TIAN Meng-qing,ZHANG Ke-jia.Pipeline Anomaly Data Detection Method Based on CNN and Bi-LSTM Network[J].,2023,33(10):215.[doi:10. 3969 / j. issn. 1673-629X. 2023. 06. 032]
[7]杨松涛,葛永琪*,王静,等.基于ME-BiLSTM模型的苜蓿叶面积指数预测方法[J].计算机技术与发展,2024,34(05):183.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0058]
　YANG Song-tao,GE Yong-qi*,WANG Jing,et al.Prediction Method of Alfalfa Leaf Area Index Based on ME-BiLSTM Model[J].,2024,34(10):183.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0058]
[8]陈君,郭立颖*,赵小会,等.基于MPBiLSTM的短期光伏发电功率预测[J].计算机技术与发展,2024,34(10):186.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0204]
　CHEN Jun,GUO Li-ying*,ZHAO Xiao-hui,et al.Short-term Photovoltaic Power Prediction Based on MPBiLSTM[J].,2024,34(10):186.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0204]
[9]史志强,古丽米拉·克孜尔别克,韩博,等.基于改进BiLSTM-KF的WSN数据漂移盲校准算法[J].计算机技术与发展,2025,(02):122.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0320]
　SHI Zhi-qiang,GULIMILA Kezierbieke,HAN Bo,et al.WSN Data Drift Blind Calibration Algorithm Based on Improved BiLSTM-KF[J].,2025,(10):122.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0320]
[10]李柄军,陈帅良,段晓英,等.基于改进的GAN与DL融合的入侵检测方法[J].计算机技术与发展,2025,(05):67.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0409]
　LI Bing-jun,CHEN Shuai-liang,DUAN Xiao-ying,et al.Intrusion Detection Method Based on Improved GAN and DL Fusion[J].,2025,(10):67.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0409]

更新日期/Last Update: 2025-10-10