XML Document Mining Using Contextual Self-organizing Maps for Structures

Kc, Milly; Hagenbuchner, Markus; Tsoi, Ah Chung; Scarselli, Franco; Sperduti, Alessandro; Gori, Marco

doi:10.1007/978-3-540-73888-6_47

Cited by 7 publications

(7 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There exist other methods such as approaches of [16,18,22,23,26,46,49,51] originally dedicated for classifying or clustering XML documents using both structure and content. These methods are in fact more general since they offer flexible models that can easily be adapted for dealing with structure only.…”

Section: Related Workmentioning

confidence: 99%

Using structural similarity for clustering XML documents

et al. 2011

View full text Add to dashboard Cite

In this paper, we describe a method for clustering XML documents. Its goal is to group documents sharing similar structures. Our approach is two-step. We first automatically extract the structure from each XML document to be classified. This extracted structure is then used as a representation model to classify the corresponding XML document. The idea behind the clustering is that if XML documents share similar structures, they are more likely to correspond to the structural part of the same query. Finally, for the experimentation purpose, we tested our algorithms on both real (ACM SIGMOD Record corpus) and synthetic data. The results clearly demonstrate the interest of our approach.

show abstract

Section: Related Workmentioning

confidence: 99%

Using structural similarity for clustering XML documents

et al. 2011

View full text Add to dashboard Cite

show abstract

“…There are also various work performed on unsupervised machine learning methods [5,6] using Self- organizing Map for Structured Data (SOM-SD) which can be used for clustering web pages. To the best of our knowledge, there is no other published work on machine learning for general web page ranking issues.…”

Section: Graph Neural Network and Related Workmentioning

confidence: 99%

Ranking Web Pages Using Machine Learning Approaches

Yong

Hagenbuchner

Tsoi

2008

2008 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology

View full text Add to dashboard Cite

“…This is because objects in computer science are popularly described by various types of parsing trees (Fu, 1977). The SOM-SD has been very successful in a range of application domains and has set the state-of-the-art performances (Hagenbuchner et al, 2006;Kc et al, 2007) in this type of problems.…”

Section: The Self Organizing Map For Structured Datamentioning

confidence: 99%

“…The traditional approach to processing data vectors in machine learning is through Multi-Layer Perceptron Neural networks (MLP) (Haykin, 1994), and Self-Organizing Maps (SOMs) (Haykin, 1994), whereas the traditional approach to processing sequences is by using Elman Networks (Elman, 1990;Kohonen, 1984;. Machine learning approaches for the processing of graphs have only recently been introduced (Hagenbuchner et al, 2003;Kc et al, 2010;Scarselli et al, 2009a), and have already been very successful in solving several benchmark problems (Hagenbuchner et al, 2006;Kc et al, 2007;Zhang et al, 2009). This chapter describes a general framework in which graph structured information can be encoded for machine learning applications.…”

Section: Introductionmentioning

confidence: 99%

“…For example, the SOM-SD holds the state-of-the-art performance for several benchmark problems (provided by INEX) on the clustering of relatively large sets of XML documents (Hagenbuchner et al, 2006;Kc et al, 2007), and has been successfully used as a filter on a spam detection task in hyper-linked domain (Di Noi et al, 2010). The versatility of the method underlines the benefits of data driven approaches, and shows that SOM based methods continue to be beneficial for solving real world data mining problems.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Data Driven Encoding of Structures and Link Predictions in Large XML Document Collections

Hagenbuchner

Tsoi

et al.

Advances in Data Mining and Database Management

View full text Add to dashboard Cite

In recent years there have been some significant research towards the ability of processing related data, particularly the relatedness among atomic elements in a structure with those in another structure. A number of approaches have been developed with various degrees of success. This chapter provides an overview of machine learning approaches for the encoding of related atomic elements in one structure with those in other structures. The chapter briefly reviews a number of unsupervised approaches for such data structures which can be used for solving generic classification, regression, and clustering problems. We will apply this approach to a particularly interesting and challenging problem: The prediction of both the number and their locations of the in-links and out-links of a set of XML documents. In this problem, we are given a set of XML pages, which may represent web pages on the Internet, with in-links and out-links. Based on this training dataset, we wish to predict the number and locations of in-links and out-links of a set of XML documents, which are as yet not linked to other existing XML documents. To the best of our knowledge, this is the only known data driven unsupervised machine learning approach for the prediction of in-links and out-links of XML documents. ABSTRACTIn recent years there have been some significant research towards the ability of processing related data, particularly the relatedness among atomic elements in a structure with those in another structure. A number of approaches have been developed with various degrees of success. This chapter provides an overview of machine learning approaches for the encoding of related atomic elements in one structure with those in other structures. The chapter briefly reviews a number of unsupervised approaches for such data structures which can be used for solving generic classification, regression, and clustering problems. We will apply this approach to a particularly interesting and challenging problem: The prediction of both the number and their locations of the in-links and out-links of a set of XML documents. In this problem, we are given a set of XML pages, which may represent web pages on the Internet, with in-links and out-links. Based on this training dataset, we wish to predict the number and locations of in-links and out-links of a set of XML documents, which are as yet not linked to other existing XML documents. To the best of our knowledge, this is the only known data driven unsupervised machine learning approach for the prediction of in-links and out-links of XML documents.

show abstract

XML Document Mining Using Contextual Self-organizing Maps for Structures

Cited by 7 publications

References 4 publications

Using structural similarity for clustering XML documents

Using structural similarity for clustering XML documents

Ranking Web Pages Using Machine Learning Approaches

Data Driven Encoding of Structures and Link Predictions in Large XML Document Collections

Contact Info

Product

Resources

About