Research on Chinese Short Text Clustering
Ensemble via Convolutional Neural Networks
Haowen Wan, Bo Ning
(&) , Xiaoyu Tao, and Jianfei Long
Dalian Maritime University, Dalian 116026, China
ningbo@dlmu.edu.cn
Abstract. Different from traditional text, short texts are characterized by high
dimensionality, sparseness, and large text size. At the same time, some existing
clustering ensemble algorithms treat each clusters equally, which will lead to the
problem that the clustering results are not good enough. To solve this problem,
this paper proposed a short text clustering ensemble algorithm based on convolution neural network (CNN). Firstly, the word2vec model is used to preserve
the semantic relationship between words and obtain the multi-dimensional word
vector representation; secondly, the feature is extracted from the original vector
combining with the CNN; thirdly, clustering methods are used to cluster vectors;
and then finally, Gini coefficient is used to measure the reliability of clustering,
and the final clustering ensemble is carried out.
Keywords: Short text Á CNN Á Clustering ensemble
1 Introduction
Text clustering has always been a classical problem in the field of text mining. It is also
the basis of natural language processing applications such as recommendation algorithms. A text will be treated as the short text when its vocabulary is less than 140
words. It is different from the traditional text. The vocabulary of short text is very few,
many words even appear only once, and so it has the problem of sparsity. Traditional
short text clustering methods such as vector space model [1] cannot solve the problems
caused by high-dimensional and sparse features. Some researchers work on expanding
and enriching the context of data from Wikipedia [2] or an ontology [3]. Some
researchers have proposed short text clustering based on semantics [4, 5]. But these
algorithms have some subjective choices and judgments more or less in vector representation, and the result is still high-dimensional word vector representation. With the
development of deep learning in recent years, convolutional neural network (CNN) is
widely used in image feature extraction, short text classification, and other fields, but it
is seldom used in short text clustering process.
Various clustering algorithms have their own advantages and disadvantages. Few
single clustering algorithms can be applied to any structured data. Compared with
clustering, clustering ensemble combines multi-base clustering to obtain better clustering results, so clustering ensemble has attracted more and more people in recent
years [6–10]. However, many existing clustering ensemble methods do not consider the
reliability of each base clustering, which leads to the low quality of the final clustering
© Springer Nature Singapore Pte Ltd. 2020
Q. Liang et al. (Eds.): Artificial Intelligence in China, LNEE 572, pp. 622–628, 2020.
https://doi.org/10.1007/978-981-15-0187-6_74
Ensemble via Convolutional Neural Networks
Haowen Wan, Bo Ning
(&) , Xiaoyu Tao, and Jianfei Long
Dalian Maritime University, Dalian 116026, China
ningbo@dlmu.edu.cn
Abstract. Different from traditional text, short texts are characterized by high
dimensionality, sparseness, and large text size. At the same time, some existing
clustering ensemble algorithms treat each clusters equally, which will lead to the
problem that the clustering results are not good enough. To solve this problem,
this paper proposed a short text clustering ensemble algorithm based on convolution neural network (CNN). Firstly, the word2vec model is used to preserve
the semantic relationship between words and obtain the multi-dimensional word
vector representation; secondly, the feature is extracted from the original vector
combining with the CNN; thirdly, clustering methods are used to cluster vectors;
and then finally, Gini coefficient is used to measure the reliability of clustering,
and the final clustering ensemble is carried out.
Keywords: Short text Á CNN Á Clustering ensemble
1 Introduction
Text clustering has always been a classical problem in the field of text mining. It is also
the basis of natural language processing applications such as recommendation algorithms. A text will be treated as the short text when its vocabulary is less than 140
words. It is different from the traditional text. The vocabulary of short text is very few,
many words even appear only once, and so it has the problem of sparsity. Traditional
short text clustering methods such as vector space model [1] cannot solve the problems
caused by high-dimensional and sparse features. Some researchers work on expanding
and enriching the context of data from Wikipedia [2] or an ontology [3]. Some
researchers have proposed short text clustering based on semantics [4, 5]. But these
algorithms have some subjective choices and judgments more or less in vector representation, and the result is still high-dimensional word vector representation. With the
development of deep learning in recent years, convolutional neural network (CNN) is
widely used in image feature extraction, short text classification, and other fields, but it
is seldom used in short text clustering process.
Various clustering algorithms have their own advantages and disadvantages. Few
single clustering algorithms can be applied to any structured data. Compared with
clustering, clustering ensemble combines multi-base clustering to obtain better clustering results, so clustering ensemble has attracted more and more people in recent
years [6–10]. However, many existing clustering ensemble methods do not consider the
reliability of each base clustering, which leads to the low quality of the final clustering
© Springer Nature Singapore Pte Ltd. 2020
Q. Liang et al. (Eds.): Artificial Intelligence in China, LNEE 572, pp. 622–628, 2020.
https://doi.org/10.1007/978-981-15-0187-6_74