Based On A Summary Of The Semantic Relation Extraction

With the development of network, there are more and more electronic texts. The abstract generated by machine is better than that of person. Abstract extraction is more and more important for us. Most of the previous automatic abstract methods are based on word counting, which misses deep semantic analysis of texts, so the generated abstract is unsatisfying.In order to overcoming the disadvantage, we put forward a new automatic abstract method based on the analysis of semantic relation and the trad auto abstract method. The new method can deal with all type texts. It takes advantage with the semantic relation of WordNet and HowNet to create the hierarchy of words. During the course of creation, we use the modified parameters to compute the importance of concept and optimize the hierarchy. At last, in order to unify the word form, we replace words with key concepts. Substitute can make all words having same form, which increases the veracity of Vector Space Model (VSM). On the basis of semantic relation abstract extraction method integrates the randomicity of processing texts with good quality.On the basis of semantic relation keywords extraction and existed two text structure analysis methods, we put forward a new method-theme partition on the basis of the degree of overlap between paragraphs. This new method integrates the two method's excellence. It has enhanced the precision rate. Above all it agrees to that instance that content is similar but word is different.