Research On The Method Of Interpreting Complicated Voice Instruction Instruction For Home Service Robot

In the field of home robot research,how to translate the voice instruction correctly into the task sequence of the robot has become one of the important research topics.The ability to parse instructions becomes a prerequisite for the robot to provide better services to humans.Based on a comprehensive analysis of the status quo of research at home and abroad,this dissertation presents the problem of complex instruction parsing for family service environment.And through the use of natural language processing,knowledge base and a variety of statistical theory,the above-mentioned issues in-depth study.Firstly,This dissertation elaborates the theory of word segmentation and the construction of grammar tree,which provides a theoretical basis for the study of the subsequent complex instruction.According to the characteristics of Chinese expression and the relevant information,four kinds of logical relations,namely,parallelism,progress,transition and conditions,are used in daily life,and the detailed analysis of these four logical relations is carried out.Secondly,after the instruction sequence is segmented and constructed into a syntax tree,the correct decomposition algorithm is designed according to the characteristics of the generated syntax tree.Through the algorithm,a complex instruction can be decomposed into two or more atomic instruction sets,and the atomic instruction is extracted.The relationship between the characteristics is extracted.In the implementation of the decomposition algorithm,it is also necessary for the instruction to complement the attributive and pronoun digestion operations to ensure that the decomposition of the instruction without affecting the meaning of the original instruction can be the case of the correct implementation.Thirdly,after analysis of the various atomic instruction sets,a named entity recognition method based on preprocessing is proposed.At present,the method of naming entity recognition in the field of family service has too low recognition rate when dealing with restrictive attributive directives.After studying the structure of attributive,especially the restrictive attributive,this dissertation proposes a preprocessing method which combines the automata to eliminate the restrictive attributive from the instruction sequence,Perform named entity identification.This method can greatly improve the accuracy of entity recognition.Finally,the correctness and validity of the proposed method and the method of named entity recognition based on preprocessing are verified by experiment.The experimental results are analyzed.