基于概率语义分布的短文本分类

马成龙; 颜永红

doi:10.16383/j.aas.2016.c150268

基于概率语义分布的短文本分类

doi: 10.16383/j.aas.2016.c150268

马成龙^1, ,,
颜永红^1,2,

1.
中国科学院声学研究所语言声学与内容理解重点实验室北京 100190
2.
新疆民族语音语言信息处理实验室乌鲁木齐 830011

基金项目:

国家重点基础研究发展计划（973计划） 2013CB329302

国家高技术研究发展计划（863计划） 2015AA016306

新疆维吾尔自治区科技重大专项 201230118-3

中国科学院战略性先导科技专项 XDA06030100, XDA06030500, XDA06040603

国家自然科学基金 11461141004, 61271426, 11504406, 11590770, 11590771, 11590772, 11590773, 11590774

详细信息

作者简介:
颜永红中国科学院声学研究所语言声学与内容理解重点实验室教授.1990年在清华大学获得学士学位, 1995年8月于美国俄勒冈研究院(Oregon GraduateInstitute, OGI) 获得计算机科学和工程博士学位.他曾在OGI 担任助理教授(1995年), 副教授(1998年) 和副主任(1997年).主要研究方向为语音处理和识别, 语言/说话人识别和人机界面.E-mail:yanyonghong@hccl.ioa.ac.cn

通讯作者:
马成龙中国科学院声学研究所博士研究生.2011年获得山东大学(威海) 通信工程学士学位.主要研究方向为自然语言处理, 口语理解, 情感分析, 深度学习.E-mail:machenglong@hccl.ioa.ac.cn

计量
- 文章访问数: 2286
- HTML全文浏览量: 199
- PDF下载量: 903
- 被引次数: 0
出版历程
- 收稿日期: 2015-05-19
- 录用日期: 2016-05-03
- 刊出日期: 2016-11-01

Short Text Classification Based on Probabilistic Semantic Distribution

MA Cheng-Long^{1
, ,},
YAN Yong-Hong^{1,2
,}

1.
The Key Laboratory of Speech Acoustics and Content Understanding, Institute of Acoustics, Chinese Academy of Sciences, Beijing 100190
2.
Xinjiang Laboratory of Minority Speech and Language Information Processing, Urumchi 830011

Funds:

National Basic Research Program of China (973 Program) 2013CB329302

National High Technology Research Program of China (863 Program) 2015AA016306

the Key Science and Technology Project of the Xinjiang Uygur Autonomous Region 201230118-3

the Strategic Priority Research Program of the Chinese Academy of Sciences XDA06030100, XDA06030500, XDA06040603

National Natural Science Foundation of China 11461141004, 61271426, 11504406, 11590770, 11590771, 11590772, 11590773, 11590774

More Information

Author Bio:
Professor at The Key Laboratory of Speech Acoustics and Content Understanding, Chinese Academy of Sciences. He received his bachelor degree from Tsinghua University in 1990, and Ph.,D. degree from Oregon Graduate Institute (OGI), USA. He worked in OGI as assistant professor (1995), associate professor (1998) and associate director (1997) of Center for Spoken Language Understanding. His research interest covers speech processing and recognition, language/speaker recognition, and human computer interface.

Corresponding author: MA Cheng-Long Ph.,D. candidate at the Institute of Acoustics, Chinese Academy of Sciences. He received his bachelor degree from Shandong University, Weihai in 2011. His research interest covers natural language processing, spoken language understanding, sentiment analysis and deep learning. Corresponding author of this paper.

摘要

摘要: 在短文本分类中，面对特征稀疏的短文本，如何充分利用文本中的每一个词语成为关键.本文提出概率语义分布模型的思想，首先通过查询词矢量词典，将文本转换为词矢量数据；其次，在概率语义分布模型的假设下利用混合高斯模型对无标注的文本数据进行通用背景语义模型训练；利用训练数据对通用模型进行自适应得到各个领域的目标领域语义分布模型；最后，在测试过程中，计算短文本属于领域模型的概率，得到最终的分类结果.实验结果表明，本文提出的方法能够从一定程度上利用短文本所提供的信息，有效降低了对训练数据的依赖性，相比于支持向量机（Support vector machine，SVM）和最大熵分类方法性能相对提高了17.7%.
- 短文本分类 /
- 词矢量 /
- 语义分布 /
- 高斯混合模型
Abstract: In short text classification, it is critical to deal with each word because of data sparsity. In this paper, we present a novel probabilistic semantic distribution model. Firstly, words are transformed to vectors by looking up word embeddings. Secondly, the universal background semantic model is trained based on unlabelled universal data through mixture Gaussian models. Then, target models are obtained by adapting the background model for each domain training data. Finally, the probability of the test data belonging to each target model is calculated. Experimental results demonstrate that our approach can make best use of each word and effectively reduce the influence of training data size. In comparison with the methods of support vector machine (SVM) and MaxEnt, the proposed method gains a 17.7% relative accuracy improvement.
- Short text classification /
- word embedding /
- semantic distribution /
- Gaussian mixture model

HTML全文

图 1 基于通用语义背景模型的短文本分类

Fig. 1 Short text classification based on universal semantic background model

下载: 全尺寸图片幻灯片

图 2 不同的背景数据和高斯数对分类结果的影响

Fig. 2 Influence of background data and \\the number of GMM

下载: 全尺寸图片幻灯片

图 3 训练数据大小对分类效果的影响(1)

Fig. 3 Influence of training set size (1)

下载: 全尺寸图片幻灯片

图 4 训练数据大小对分类效果的影响(2)

Fig. 4 Influence of training set size (2)

下载: 全尺寸图片幻灯片

表 1 网页搜索片段数据分布

Table 1 Statistics of web snippets data

编号	领域	训练数据	测试数据
1	商业	1200	300
2	计算机	1200	300
3	文化与艺术	1880	330
4	教育与科技	2360	300
5	技术	220	150
6	健康	880	300
7	社会政策	1200	300
8	体育	1120	300
共计		10060	2280

下载: 导出CSV

表 2 未登录词分布

Table 2 Statistics of unseen words

	原始单词	词干
训练数据	26 265	21 596
测试数据	10 037	8 200
未登录词	4 378	3 677
未登录词的比例	43.62%	44.84%

下载: 导出CSV

表 3 与基线系统对比实验结果(%)

Table 3 Experimental results of the proposed method against other methods (%)

方法	Accuracy
TF*IDF+SVM	66.14
TF*IDF+MaxEnt	66.80
LDA+MaxEnt	82.18
Wiki feature+SVM	76.89
Paragraph vector+SVM	61.90
LSTM	63.00
本文的方法	80.00

下载: 导出CSV

表 4 SVM、MaxEnt和本文方法的实验结果

Table 4 Evaluations of SVM,MaxEnt and the proposed method

		SVM			MaxEnt			本文的方法
领域	P (%)	R (%)	F1	P (%)	R (%)	F1	P (%)	R (%)	F1
社会政策	77.61	52.00	0.6228	70.75	50.00	0.5859	86.36	70.37	0.7755
计算机	73.75	63.67	0.6834	72.26	66.00	0.6899	80.31	87.29	0.8365
教育与科技	41.98	82.00	0.5553	45.93	82.67	0.5905	81.60	68.23	0.7432
体育	85.19	76.67	0.8070	86.08	78.33	0.8202	84.54	89.93	0.8715
健康	89.01	56.67	0.6925	86.94	64.33	0.7395	76.35	85.57	0.8070
技术	76.53	50.00	0.6048	72.84	39.33	0.5108	58.82	93.33	0.7216
商业	70.37	57.00	0.6298	68.05	60.33	0.6396	73.99	67.33	0.7051
文化与艺术	62.27	81.52	0.7060	62.86	78.48	0.6981	88.15	77.85	0.8268

下载: 导出CSV

参考文献(24)

[1]	Wang B K, Huang Y F, Yang W X, Li X. Short text classification based on strong feature thesaurus. Journal of Zhejiang University Science C, 2012, 13(9): 649-659 doi: 10.1631/jzus.C1100373
[2]	Zelikovitz S, Hirsh H. Improving short text classification using unlabeled background knowledge to assess document similarity. In: Proceedings of the 17th International Conference on Machine Learning. San Francisco, USA: Morgan Kaufmann, 2000. 1183-1190
[3]	Bollegala D, Matsuo Y, Ishizuka M. Measuring semantic similarity between words using web search engines. In: Proceedings of the 16th International Conference on World Wide Web. New York, USA: ACM, 2007. 757-766
[4]	Gabrilovich E, Markovitch S. Computing semantic relatedness using Wikipedia-based explicit semantic analysis. In: Proceedings of the 20th International Joint Conference on Artificial Intelligence. San Francisco, USA: Morgan Kaufmann, 2007. 1606-1611
[5]	Banerjee S, Ramanathan K, Gupta A. Clustering short texts using Wikipedia. In: Proceedings of the 30th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval. New York, USA: ACM, 2007. 787-788
[6]	Lucene [Online], available: https://lucene.apache.org/, May 3, 2016.
[7]	Phan X H, Nguyen L M, Horiguchi S. Learning to classify short and sparse text & web with hidden topics from large-scale data collections. In: Proceedings of the 17th International Conference on World Wide Web. New York, USA: ACM, 2008. 91-100
[8]	Blei D M, Ng A Y, Jordan M I. Latent Dirichlet allocation. Journal of Machine Learning Research, 2003, 3: 993-1022 http://cn.bing.com/academic/profile?id=1880262756&encoded=0&v=paper_preview&mkt=zh-cn
[9]	Kim Y. Convolutional neural networks for sentence classification. In: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. Doha, Qatar: Association for Computational Linguistics, 2014. 1746-1751
[10]	Le Q, Mikolov T. Distributed representations of sentences and documents. In: Proceedings of the 31st International Conference on Machine Learning. Beijing, China: JMLR, 2014. 1188-1196
[11]	Kalchbrenner N, Grefenstette E, Blunsom P. A convolutional neural network for modelling sentences. In: Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics. Baltimore, USA: Association for Computational Linguistics, 2014. 655-665
[12]	Landauer T K, Foltz P W, Laham D. An introduction to latent semantic analysis. Discourse Processes, 1998, 25(2-3): 259-284 doi: 10.1080/01638539809545028
[13]	Mikolov T, Chen K, Corrado G, Dean J. Efficient estimation of word representations in vector space. arXiv: 1301.3781, 2013.
[14]	Turian J, Ratinov L, Bengio Y. Word representations: a simple and general method for semi-supervised learning. In: Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics. Uppsala, Sweden: Association for Computational Linguistics, 2010. 384-394
[15]	Mikolov T, Yih W T, Zweig G. Linguistic Regularities in Continuous Space Word Representations. In: Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Atlanta, Georgia: Association for Computational Linguistics, 2013. 746-751
[16]	Reynolds D A. Speaker identification and verification using Gaussian mixture speaker models. Speech Communication, 1995, 17(1-2): 91-108 doi: 10.1016/0167-6393(95)00009-D
[17]	Reynolds D A, Quatieri T F, Dunn R B. Speaker verification using adapted Gaussian mixture models. Digital Signal Processing, 2000, 10(1-3): 19-41 doi: 10.1006/dspr.1999.0361
[18]	Collobert R, Weston J, Bottou L, Karlen M, Kavukcuoglu K, Kuksa P. Natural language processing (almost) from scratch. Journal of Machine Learning Research, 2011, 12: 2493-2537
[19]	Mikolov T, Sutskever I, Chen K, Corrado G S, Dean J. Distributed representations of words and phrases and their compositionality. In: Proceedings of the 2013 Advances in Neural Information Processing Systems. Lake Tahoe, Nevada, USA: Curran Associates, Inc., 2013. 3111-3119
[20]	Porter M F. An algorithm for suffix stripping. Readings in Information Retrieval. San Francisco: Morgan Kaufmann, 1997. 313-316
[21]	Ling G C, Asahara M, Matsumoto Y. Chinese unknown word identification using character-based tagging and chunking. In: Proceedings of the 41st Annual Meeting on Association for Computational Linguistics. Sapporo, Japan: Association for Computational Linguistics, 2003. 197-200
[22]	Parker R, Graff D, Kong J B, Chen K, Maeda K. English Gigaword Fifth Edition [Online], available: https://catalog.ldc.upenn.edu/LDC2011T07, May 3, 2016.
[23]	Wang P, Xu B, Xu J M, Tian G H, Liu C L, Hao H W. Semantic expansion using word embedding clustering and convolutional neural network for improving short text classification. Neurocomputing, 2016, 174: 806-814 doi: 10.1016/j.neucom.2015.09.096
[24]	Hochreiter S, Schmidhuber J. Long short-term memory. Neural Computation, 1997, 9(8): 1735-1780 doi: 10.1162/neco.1997.9.8.1735