Automatic Summarization Based on Mutual Information

Abstract:

Many existing automatic summarization methods often process a text based on the paragraphs. These methods ignore the correlation degree among paragraphs, which lead to the inaccuracy of the extracted summary. To solve the problem, we present a method based on mutual information. Firstly, we divide the whole text into several smaller blocks which relevant to each topic by measuring the degree of association among words, sentences and paragraphs. Then we compute the weight of the sentences by using the improved formula which aims to evaluate the importance of one sentence to a topic. Finally, the summary can be generated by combining the sentences selected from the different topics. The experiment result shows the validity of the method.

• Periodical: Applied Mechanics and Materials (Volumes 513-517)
• Pages: 1994-1997
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.513-517.1994
• Online since: February 2014
• Authors: Hua Huo*, Xing Han Liu
• Keywords: Automatic Abstracting, Mutual Information (MI), Topic Classification

References

[1] Salton G, Singhal A, Mitra M, et al. Automatic Text Structuring and Summarization (Information Processing and Management)Vol. 33(1997): pp.193-207, in English.

DOI: 10.1016/s0306-4573(96)00062-3

[2] Weijun Gao, Donglin Ma. An ontology-based text topic extraction method research (Computer Applications and software)Vol. 29(2012): pp.54-56, in Chinese.

[3] Qiuchan Bai, Chunxia Jin. Topic-text clustering algorithm based on word co-occurrence Vol. 35(2013): pp.164-168, in Chinese.

[4] Xinguo Lu, Yaping Lin. An Improved Feature Selection Preprocessing Algorithm Based on Mutual Information Vol. 32(2008): pp.104-107, in Chinese.

[5] David Zajic, Bonnie Dorr, Jimmy Lin, et al . A sentence-trimming approach to multi-document summarization . Proceedings of the HLT -EMNLP 2005 Works hop on Text Summarization.