Knowledge Based Summarization and Document Generation using Bayesian Network

In this paper an approach of Semantic Knowledge Extraction (SKE), from a set of research papers, is proposed to develop a system Summarized Research Article Generator (SRAG) which would generate a summarized research article based on the query given by a user. The SRAG stores the semantic knowledge extracted from the query relevant papers in the form of a semantic tree. Semantic Tree stores all the textual units with their score in nodes organized at different levels depending on their type such as at the bottom leaf nodes keep the words with its probability, the upper level of it represent sentences with its score, next to it paragraphs, segments and so on. Scores of all the entities are calculated in bottom to up manner, first score of words are calculated, based on words sentences are ranked and similarly all the higher levels of the knowledge tree would be scored. A method of Bayesian network is used to generate a probabilistic model which would extract the relevant information from the knowledge tree to generate a summarized article. To maintain coherency, the summarized document is generated segment-wise by combining the most relevant paragraphs. Abstract of a generated summary is shown as a sample result. To show the effectiveness of the algorithm, an intrinsic evaluation strategy, degree of representativeness (DOG) is used. DOG gives on average 50% of relevance of the summary with the source. It's been observed that the proposed approach generates a comprehensive and precise papers.