基于Tri-Training的事件关系分类方法研究

J4 ›› 2015, Vol. 37 ›› Issue (12): 2345-2351.

基于Tri-Training的事件关系分类方法研究

丁思远，洪宇，朱珊珊，姚建民，朱巧明

(苏州大学江苏省计算机信息处理重点实验室，江苏苏州 215006)

收稿日期:2015-09-01 修回日期:2015-11-03 出版日期:2015-12-25 发布日期:2015-12-25
基金资助:
国家自然科学基金资助项目(61003152,61272259,61272260)

Event relation classification based on Tri-Training

DING Siyuan,HONG Yu,ZHU Shanshan,YAO Jianmin,ZHU Qiaoming

(Provincial Key Laboratory of Computer Information Processing Technology,Soochow University,Suzhou 215006,China)

Received:2015-09-01 Revised:2015-11-03 Online:2015-12-25 Published:2015-12-25

摘要/Abstract

摘要：

事件关系分类是一项研究事件之间存在何种逻辑关系的自然语言处理技术。针对事件关系分类任务中训练语料不足的问题，提出了基于Tri-Training的事件关系分类方法。该方法首先根据已标注的语料训练三个不同的分类器，以多数投票的方式从未标注集中抽取置信度较高的样本对训练集进行扩充，然后利用新的训练集重新训练分类器，反复迭代，不断完善分类模型，最终达到提升事件关系分类性能的目的。实验结果表明，以F1值为评价标准，基于Tri-Training的事件关系分类方法在四大类事件关系上的分类性能为64.36%。

关键词: 事件关系, 框架语义, 半监督学习, Tri-Training

Abstract:

As one of natural language processing techniques, event relation detection aims at exploring logical relationship between pairwise events. To solve the problem of lacking enough training data in event relation detection tasks, we propose a novel approach based on Tri-Training to augment the training corpus. We firstly use labeled training data to learn three different classifiers, and then exploit majority voting method to expand training corpus with higher confidence samples, iteratively optimize the model and eventually improve the performance of event relation classification. Experimental results show that compared to other methods, the TriTraining based method achieves 64.3% F1-score over four general semantic relations.

Key words: event relation;frame;semi-supervised learning;Tri-Training

丁思远，洪宇，朱珊珊，姚建民，朱巧明. 基于Tri-Training的事件关系分类方法研究[J]. J4, 2015, 37(12): 2345-2351.

DING Siyuan,HONG Yu,ZHU Shanshan,YAO Jianmin,ZHU Qiaoming. Event relation classification based on Tri-Training [J]. J4, 2015, 37(12): 2345-2351.

参考文献

［1］Yang X R, Hong Y, Chen Y D, et al.Detecting event relation through crossscenario Inference ［J］.Journal of Chinese information Processing,2014,28(5):206214.(in Chinese)
［2］Fillmore C J,Johnson C R,Petruck M R L.Background to framenet［J］.International Journal of Lexicography,2003,16(3):235250.
［3］Hashimoto C,Torisawa K,Kloetzer J,et al.Toward future scenario generation:Extracting event causality exploiting semantic relation,context,and association features［C］∥Proc of the 52nd Annual Meeting of the Association for Computational Linguistics,Association for Computational Linguistics,2014:1.
［4］Chklovski T,Pantel P.Global pathbased refinement of noisy graphs applied to verb semantics［C］∥Natural Language ProcessingIJCNLP 2005,2005:792803.
［5］Pantel P,Pennacchiotti M.Espresso:Leveraging generic patterns for automatically harvesting semantic relations［C］∥Proc of the 21st International Conference on Computational Linguistics and the 44th Annual Meeting of the Association for Computational Linguistics.Association for Computational Linguistics,2006:113120.
［6］Harris Z S.Mathematical structures of language［M］.New York:Interscience Publishers John Wiley & Sons,1968.
［7］Lin D,Pantel P.DIRT@ SBT@ discovery of inference rules from text［C］∥Proc of the 7th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining,2001:323328.
［8］Zhou Z H,Li M.Tritraining:Exploiting unlabeled data using three classifiers［J］.IEEE Transactions on Knowledge and Data Engineering,2005,17(11):15291541.
［9］Goldman S,Zhou Y.Enhancing supervised learning with unlabeled data［C］∥Proc of ICML,2000:327334.
［10］Linguistic Data Consortium.ACE (Automatic Content Extraction) Chinese annotation guidelines for events［EB/OL］.［20090908］.http://www.ldc.upenn.edu/Projects/ACE.
［11］Li Q,Ji H,Huang L.Joint event extraction via structured prediction with global features［C］∥Proc of the 51st Annual Meeting of the Association for Computational Linguistics,2013:7382.
［12］Zhu X.Semisupervised learning literature survey［R］.Technical Report 1530. Madison:Computer Sciences, Univerity of Wisconsinnadison,2005.
附中文参考文献：
［1］杨雪蓉,洪宇,陈亚东,等.基于跨场景推理的事件关系检测方法［J］.中文信息学报,2014,28(5):206214.

[1]	王栋, 杨珂, 玄佳兴, 韩雨桐, 赵丽花, 王旭仁. 基于半监督生成对抗网络的恶意代码家族分类实现[J]. 计算机工程与科学, 2022, 44(05): 826-833.
[2]	李方, 吴国栋, 涂立静, 刘玉良, 查志康, 李景霞. 图自编码器推荐研究综述[J]. 计算机工程与科学, 2022, 44(02): 335-344.
[3]	王保成, 刘利军, 黄青松, . 基于LDA和卷积神经网络的半监督图像标注方法[J]. 计算机工程与科学, 2022, 44(01): 110-117.
[4]	孙庞博, 符琦, 陈安华, 蒋云霞. 基于组合预测模型的小样本轴承故障分类诊断 [J]. 计算机工程与科学, 2021, 43(09): 1684-1691.
[5]	黄一龙，李培峰，朱巧明. 中文事件相关性语料库构建及识别方法[J]. J4, 2015, 37(12): 2306-2311.
[6]	王晓栋,严菲,谢勇,江慧琴. 基于稀疏图表示的特征选择方法研究[J]. J4, 2015, 37(12): 2372-2378.
[7]	彭波,徐天伟,李臻,高炜. 迭代拉普拉斯半监督学习本体算法[J]. J4, 2014, 36(11): 2164-2168.
[8]	袁兴梅,谢雪莲. 基于RSC模型和噪声去除的半监督训练方法[J]. J4, 2013, 35(9): 162-166.
[9]	汪春亮1，2,伏玉琛2. 一种针对非平衡数据的贝叶斯分类算法[J]. J4, 2010, 32(7): 95-98.
[10]	何振峰. 一种两阶段的限制层次聚类算法[J]. J4, 2006, 28(7): 70-72.