A Study Of Classification Method About Mechanical Design Scheme Based On Characteristics

Classification is an important branch of natural language understanding, as used herein means a combination of knowledge and statistics for the text is concerned, respectively after word processing, feature extraction and selection of words, the classification of said evaluation of the classification results were discussed. Achieving the purposes of the analysis and understanding of the need, providing guidance for the subsequent design advice.Firstly, the paper introduces the development of natural language understanding and overview of the classification of the situation. Then, they elaborated a common knowledge that the word feature extraction, classification, classification performance evaluation method for deployment later laid a solid foundation; then, taking into account the application of the conventional method to extract feature words after the word has many drawbacks, many features can be found in word form chunks, so not only can reduce the dimension of the text, but you can make out of the feature words extracted more representative. Then, after treatment; this paper on the "single string", "synthetic", "The affix type" three cases discussed, by disabling the word matching, affixes, such as chunks of screening, which is after the word processing word text extraction feature words, this paper based on the mean and variance of extracting feature words method; Next, combine the extracted feature words of the text to be classified, weights are calculated for each feature of the word, to weed out the smaller weights feature words, with the last-word retains the characteristics of the text to be classified vector; later, the spirit of the idea of the concept of classification, based on the words in front of the extracted features build concept hierarchy tree, when a new concept appears, according to property constraints to achieve a new concept concept hierarchy tree location, thus achieve the purpose of classification; and finally, to evaluate the effect of classification and evaluation methods in this paper are based on semantic similarity, semantic good points by calculating the previously known classes of text and each category of similar degree, get a result, according to the results of this sort on the front of the classification of the quality evaluation. The whole process of classification is a combination of knowledge and statistics, collected a large number of examples in this article, the statistics knowledge regularity. In the end, summed up a setof classification methods, this method is applied to the design of mechanical transmission, sorting out the whole process of mechanical transmission design classification established concept hierarchy tree of knowledge and the art of design in the field of mechanical transmission library. According to the needs of end-mechanical transmission screened better mechanical transmission design of programs to improve the design efficiency of the product.