Application of Machine Learning Techniques in Persian Text Summarization Systems

With the rapid growth of the Internet and the increasing information and online resources to find the desired information from large volumes of information to users is difficult. The method proposed in this paper, a system based on neural network is summarized text maker. In the proposed system, the original sentences based on features such as similarity with title and similarities with the center of the text, sentence length, sentence position, positive and negative words, valuable keywords scores receive. These points can be extracted from the sentences in a training body. According to the training corpus of sentences, in summaries to be or not to be labeled manually. Neural Network Based on these corpus, trained and finally used in the neural network by receiving scores of sentences extracted from a test text, decides whether it should be included in the final summaries. Experimental results on a documents of database Hamshahri show that the proposed system is able to with the scale F about 0.67 summaries do.