Research On The Generation Of Traffic Information Standard ICS Questionnaire Based On Deep Learning

Standard conformance test is to test the target system according to the special provisions required by a standard,so as to determine the degree of conformance between the system and the standard,which plays an important role in improving the standardization of the system.The starting point of conformance testing is the standard conformance clause and the realization of conformance statement.The conformance clause needs to be extracted from the standard clause,and the realization of conformance statement can be carried out by ICS questionnaire.Traditionally,these two works are basically completed by manpower,which is time-consuming and laborious.Therefore,this paper proposes the standard terms classification algorithm and ICS questionnaire automatic generation algorithm,and develops a test system to improve the efficiency of consistency testing.The research content of this paper is as follows:First of all,in order to automatically obtain the conformance clauses in transportation information standards,combining the bidirectional long short-term memory(BLSTM)based text enhancement representation with the CNN based sentence classification,the classification method was proposed to classify the conformance clauses for solving the problem of lack of context meaning in convolution neural network and gradient disappearance and gradient dispersion in cyclic neural network in the existing text classification methods.The core idea was to add the vectors generated by forward and backward processes of BLSTM,and then added vectors were spliced with the original vector as vector representation of the text.The text was classified as the input of CNN network.To verify the proposed model,the comparative test with the traditional machine model TF-IDF+SVM,single CNN and BLSTM neural network model and classic hybrid model was set up.According to the test of data set of standards terms of transportation information,the accuracy of the chain-mixed neural network model based on improved BLSTM and CNN reaches 93.77%.The results show that the proposed model has the best performance in the comparative experiments.Secondly,the automatic generation of ICS questionnaire is studied on the basis of consistency clause extraction.Through the analysis and research of ICS questionnaire and Chinese questions,the types of questions in the questionnaire include positive and negative questions,special questions and selective questions.On the basis of the research on the rule based,extraction based and generative text generation methods,an improved question generation model based generative adversative nets is proposed.In this paper,the model is divided into two parts: generator and discriminator.In the generator part,encoder-decoder is used to improve the model and convolution neural network is used in the discriminator.The validity of the model in question,coherence and meaning is verified by experiments.Finally,the above methods are integrated,and the system is designed,implemented and tested.The proposed algorithm is proved to be effective by using the constructed data set of transportation information standard clauses,which can promote the work of standard conformance test and improve the automation of conformance test.