基于共享背景主题的Labeled LDA模型

doi:10.3969/j.issn.0372-2112.2013.09.020

电子学报 ›› 2013, Vol. 41 ›› Issue (9): 1794-1799.DOI: 10.3969/j.issn.0372-2112.2013.09.020

基于共享背景主题的Labeled LDA模型

江雨燕, 李平, 王清

安徽工业大学管理科学与工程学院, 安徽马鞍山 243002

收稿日期:2013-01-25 修回日期:2013-06-03 出版日期:2013-09-25
- 通讯作者:
- 王清
- 作者简介:
- 江雨燕 女,1966年生于安徽宣城,安徽工业大学管理科学与工程学院副教授、硕士生导师,主要研究方向为机器学习、CSCW、信息集成;李平 男,1987年生于河北藁城, 安徽工业大学管理科学与工程学院硕士研究生,研究方向为机器学习、数据挖掘.
- 基金资助:
- 国家自然科学基金 (No.71172219); 安徽省自然科学研究项目省级重点项目 (No.KJ2011Z039,No.KJ2013A053)

Labeled LDA Model Based on Shared Background Topics

JIANG Yu-yan, LI Ping, WANG Qing

School of Management Science and Engineering, Anhui University of Technology, Ma'anshan, Anhui 243002, China

Received:2013-01-25 Revised:2013-06-03 Online:2013-09-25 Published:2013-09-25

摘要/Abstract

摘要： 隐藏狄利克雷分配(Latent Dirichlet Allocation,LDA)模型被广泛应用于文本分析、图像识别等领域.但由于LDA及其扩展模型多为无监督学习模型,无法将其应用于分类任务中.本文通过研究文档标记与LDA模型中主题的映射关系,提出一种新的Labeled LDA模型(Shared Background Topics Labeled LDA,SBTL-LDA).在SBTL-LDA模型中每个标记除了存在若干个独享的局部主题外,还存在若干个共享的背景(Background)主题,这样可以有效分析不同标记所含主题之间的依赖关系,而文档标记被映射为局部主题和共享主题的组合,因此SBTL-LDA模型可以有效提升文档标记判别的准确性.同时SBTL-LDA模型还可以看成是一种半监督聚类模型,在对文档进行聚类分析的过程中模型可以有效的利用文档的标记信息提升文档聚类效果.实验证明SBTL-LDA模型能够有效解决PLDA模型中主题之间的相似性和依赖关系,具有良好的多标记判别能力,并且具有优于LDA、PLDA模型的文档聚类效果.

关键词: 隐藏狄利克雷分配, 文本分析, 多标记学习, 半监督聚类

Abstract: LDA (Latent Dirichlet Allocation) is widely used in text analysis and images processing.However,LDA and most of its modifications are unsupervised learning models,which are not appropriate for classification especially multi-label classification problem.Through the study on the multi-label documents and LDA models,this paper proposes a new Labeled LDA model,namely Shared Background Topics Labeled LDA(SBTL-LDA).In this new model,each label has not only a set of local topics,but also has several background (global) topics.Experienmental results show that SBTL-LDA can decrease the affect of similarities and dependence between different topics and because the label of document is mapped as a combination of local topics and shared topics,so it has a high accuracy when learning from multi-Labeled documents.In addition,this model can be viewed as a semi-supervised clustering model which can utilize the information of labels and outperfom other models.

Key words: latent Dirichlet allocation, text analysis, multi-label learning, semi-supervised clustering

中图分类号:

TN911.23

江雨燕, 李平, 王清. 基于共享背景主题的Labeled LDA模型[J]. 电子学报, 2013, 41(9): 1794-1799.

JIANG Yu-yan, LI Ping, WANG Qing. Labeled LDA Model Based on Shared Background Topics[J]. Acta Electronica Sinica, 2013, 41(9): 1794-1799.

参考文献

[1] 王李冬,魏宝刚,袁杰.基于概率主题模型的文档聚类[J].电子学报,2012,11(11):2346-2350. Wang Li-dong,Wei Bao-gang,Yuan Jie.Document clustering based on probabilistic topic model[J].Acta Electronica Sinica,2012,11(11):2346-2350.(in Chinese)
[2] 吴永辉,王晓龙,丁宇新, 徐军,郭鸿志.基于主题的自适应、在线网络热点发现方法及新闻推荐系统[J].电子学报,2010,11(11):2620-2624. Wu Yong-hui,Wang Xiao-long,Ding Yu-xin,Xu Jun,Guo Hong-zhi.Adaptive on-line web topic detection method for web news recommendation system[J].Acta Electronica Sinica,2010,11(11):2620-2624.(in Chinese)
[3] Blei D M,Ng A Y,Jordan M I.Latent Dirichlet allocation[J].Machine Learning Research,2003,3:993-1022.
[4] Lafferty J D,Blei M D.Correlated topic models [A].Advances in Neural Information Processing Systems,Proceedings of the 2005 Conference [C].Vancouver:Bradford Books,2006.147-155.
[5] Li W,McCallum A.Pachinko allocation:DAG-structured mixture models of topic correlations [A].Proceedings of the 23rd International Conference on Machine Learning [C].New York:ACM,2006.577-584.
[6] D M Blei,J McAuliffe.Supervised topic models [A].Advances in Neural Information Processing System [C].Vancouver,British Columbia Canada:Curran,2008.121-128.
[7] Ramage D,Hall D,Nallapati R,et al.Labeled LDA:A supervised topic model for credit attribution in multi-labeled corpora [A].Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing Association for Computational Linguistics [C].Singapore:Springer,2009.248-256.
[8] Ramage D,Manning C D,Dumais S.Partially labeled topic models for interpretable text mining [A].Proceedings of the 17th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining [C].New York:ACM,2011.457-465.
[9] Hofmann T.Probabilistic latent semantic analysis [A].Proceedings of the Fifteenth Conference on Uncertainty in Artificial Intelligence [C].Morgan Kaufmann,San Mateo,CA:Morgan Kaufmann Publishers Inc,1999.289-296.
[10] Minka T,Lafferty J.Expectation-propagation for the generative aspect model [A].Proceedings of the Eighteenth Conference on Uncertainty in Artificial Intelligence [C].Morgan Kaufmann,San Mateo,CA:Morgan Kaufmann Publishers Inc,2002.352-359.
[11] Griffiths T L,Steyvers M.Finding scientific topics[J].National Academy of Sciences of the United States of America,2004,101(Suppl 1):5228-5235.
[12] Griffiths T L,Steyvers M,Blei D M,et al.Integrating topics and syntax[J].Advances in Neural Information Processing Systems,2005,17:537-544.
[13] Zhang M L,Zhou Z H.ML-KNN:A lazy learning approach to multi-label learning [J].Pattern Recognition,2007,40(7):2038-2048.
[14] Spyromitros E,Tsoumakas G,Vlahavas I.An empirical study of lazy multilabel classification algorithms [A].Proceedings of the 5th Hellenic Conference on Artificial Intelligence [C].Berlin,Heidelberg:Springer-Verlag,2008.401-406.
[15] Cheng W,Hüllermeier E.Combining instance-based learning and logistic regression for multilabel classification[J].Machine Learning,2009,76(2-3):211-225.
[16] C Vens,J Struyf,L Schietgat,S Dzeroski,H Blockeel.Decision trees for hierarchical multi-label classification[J].Machine Learning,2008,73(2):185-214.

基于共享背景主题的Labeled LDA模型

Labeled LDA Model Based on Shared Background Topics

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 2

编辑推荐

Metrics

[1]	朱赛赛, 贾修一, 李泽超. 一种基于全局和局部标记相关性的多标记分类算法[J]. 电子学报, 2020, 48(12): 2345-2351.
[2]	王娜;李霞. 基于监督信息特性的主动半监督谱聚类算法[J]. 电子学报, 2010, 38(1): 172-176.