Intention Understanding Of Nature Language Instruction In Human Computer Interaction

Among all kinds of human-computer interaction methods,interacting by speech is one of the most important ones,especially in natural language which is the most direct and friendly way.Intention understanding in natural language is a very important technology in speech human-computer interaction.In this paper,I study the algorithm of understanding the natural language instruction in human-computer interaction.First,a corpus of text-instructions in natural language is created according to the working environment and application scenarios for the home service robot and a intention tree is built,then each text instruction is labeled with a corresponding intention so as to transform the problem of intention understanding in natural language into a simply classification one;Secondly,on the basis of the traditional text vector space model,the part-of-speech vector space model which contains the information of part-of-speech is defined;Thirdly,stacked Denoising Autoencoder(SDAE)is applied to text-instruction intention understanding for extracting the higher level features of the text-instructions.The high order feature of the text-instruction is extracted to improve the robustness of the system and to improve the accuracy;Finally,support vector machine is used for training and prediction in order to achieve text-instruction intention understanding in natural language.I verify the effectiveness of the algorithm by multi fold cross validation on the constructed corpus.The results show that the average accuracy of instructions is more than 96%,which is higher than that of the traditional algorithm.