Research Of Intelligent Remark Generation Based On GA-BP Nerve Network

Nature language of machine processing is an important research field of computer science, with aim of research of intelligent comments, the paper tries to achieve application system which can automatically generate comments with students' characteristic by comprehensive use of educational psychology, linguistics and computer science, and eases a lot of work for head teacher. The paper's thesis derives from project of the innovation fund of science and technology by 2009 which is application research of combining theory with practice by argument on the results of computer nature language processing.Existing universal and efficient ICTCLAS word segmentation algorithm is used in the paper for word segmentation, part-of-speech tagging of the language materials and grammatical feature extraction, according to Chinese syntax characteristic, a specific remark corpus is founded which contains rule base, Chinese vocabulary library, grammar library, and custom database. in the support of remark corpus and specific remark's template, it realizes generation of intelligent comments by GA-BP nerve network and the TF-IDF textual similarity calculation which is based on the space of a model.Firstly, the paper introduces collection of remark material, establish the remark corpus and custom database of language material after the detailed grammar research. Secondly, the principle of evaluation target system is discussed in the paper, and also 11 impact factors by the psychology and character traits are selected to evaluate the performance of students. Thirdly, BP nerve network and genetic algorithm are introduced in the paper, and the feasibility of the editor of the comments is explained by the combination of genetic algorithm and BP nerve network.; Finally, Taking example of designing evaluation management and the editor of comments, which realizes the desired functions such as sample's maintenance, network training, nerve network's evaluation and fulfills the generation of personalized comments under the supporting of remark corpus with the comprehensive application of word segmentation technology and textual similarity calculation.Through experimental data, it indicates that GA-BP for students' evaluation is completely feasible with accuracy and rate of convergence is superior to BP nerve network. At the same time, the editor makes exact comments with the characters of students. The subject of research and fruit have some good reference value and application prospect in Chinese information processing of various areas, especially in educational, scientific and technological direction.