Developing A Framework For Summarizing XML Documents

Abstract

XML has become one of the default standards data representation over the World Wide Web (WWW) and elsewhere as many data are stored in this format. If we want to understand these documents with its complex in both structure and data, we must spend much time reading them. Sometimes, we can find it is impracticable, but not impossible to read the whole document in case of the document is large and more complex, so with the exchange of data, it is expected to see ever growing repositories of the XML documents. Also, it can be used in various applications as its flexibility and easy to use so the need to summarize XML document become increasingly an important topic to save time and cost. For these reasons, there is more interest in developing tools for summarizing XML Documents. There are two kinds of summaries that can be generated: (1) Generic Summarization based on the entire contents of the XML documents "A generic summary summarizes the entire contents of the XML document" [1]. (2)A Query-Biased summarization which summarizes the parts of the document which are relevant to what the user types on his query [1]. In producing generic summaries the implicit assumption there regarding to the information need for the user is that he is interested in knowing what is in the XML document, without having to read it in its entirety. In this thesis, we focus on generic summaries for XML documents. However, there are few works have developed for building a framework for summarizing XML documents this is mainly due to the lack of resources that are necessary for the development of such systems. We categorized the XML summarization approaches to Structural summarization, and Content and Structure based summarization. This thesis introduces an XML Abstractive Summary (XAS) approach to summarize text in the format of an XML document that is called XML summarization. XAS approach is considered a new attempt to produce an abstractive summary of the xml document regarding to performance and accuracy. The output document is a concise and readable version of the original one. The experiments are done using two dataset: IMDB and DBLP. The results has been tested with more than 300000 XML documents. XAS approach