Loading...

Abstractive Persian User Reviews Summarization in the Frameworks of Graph Data Structure and Complex Networks

Bazargan, Sara | 2016

1034 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 49359 (31)
  4. University: Sharif University of Technology
  5. Department: Languages and Linguistics Center
  6. Advisor(s): Khosravizadeh, Parvaneh
  7. Abstract:
  8. With the rapid development in information-communication technologies a huge amount of electronic documents has been produced and increase everyday in the world wide web. According to this wide information environment, summarization is critical for the users who digest this electronic big data. This research proposes two graph-based abstractive models for summarizing the Persian texts. A good summary should cover the overall context and the important subjects, and should be properly readable and coherent as well. Some permanent challenges in automatic text summarizations intended to increase the readability of output text, as well as covering all the main topics, and minimize the redundancy. In the suggested approaches, graph is constructed by the words of the input text and so it’s large in size. Then the output summary generated according to this graph’ properties. Recently, for studying the large size networks which are based on graphs, has been used of complex networks concepts. Since the networks are the common structures of human systems, the purpose is to implement the suggested framework with the use of extracting complex networks properties of textual graph to model the summarization system. These approaches are unsupervised and use a very shallow syntactic information. A reason for suggesting these models, on the one hand, is the increased use of hand-held devices such as mobile phones that requires sentences to be more compact. On the other hand, the importance of saving the time has highlighted the need for strategies accelerate the access to information. The goal is to generate an ultra-concise and well-formed summary using the graph/network properties. Evaluation results show that the system generated summaries have high compatibility with human generated summaries
  9. Keywords:
  10. Summarization ; Complex Network ; Text Summarization ; Abstractive Summarization ; Graph Data Structure

 Digital Object List

 Bookmark

...see more