Using a Semantic Fuzzy System to Intelligent Documents Summarization

Due to the information technology revolution, there are many and varied methods of document summarization to obtain specific information from documents. Automated summarization methods rely on identifying important points in all relevant documents to produce a concise summary. Therefore, this paper presents an intelligent classification-based automated summarization system using a semantic neuro-fuzzy approach. The proposed system consists of five integrated phases, which are the Document Pre-processing, the intermediate representation, the Index Matrices Weight Calculation, the Neuro-fuzzy system, and the Summary Generation, respectively. The first stage divides paragraphs into sentences and sentences into words, by removing the most frequent words that do not carry any information and stripping the word from suffixes and prefixes to extract the « root » of the words. In the second stage, the Latent Semantic Index was used to produce the words/concepts matrix and concepts/sentences matrix. The third stage used the pointwise mutual information measure that defines particularly informative about the target word, as well as the best weighting of association between words. The knowledge is then extracted using a neuro-fuzzy network learning technique in phase four, which encodes the learned knowledge in its structure as a set of fuzzy rules.