| With the development of the intelligent service robot, intelligent service robot has entered people’s life. Language is the most natural way of communication, using voice to guide the robot to human services more and more get of people love. Deep voice information and surface information have been identified, through careful study find that there is a certain relationship between the surface and deep information, such as the inclusion relation, timing relationships, etc, and put the problems into entity relation extraction. In this article, intelligent service robot is carried out on the understanding of Chinese instruction key information. We put forward a method based on named entity recognition to build relationship knowledge base, which to help service robots to understand instructions and extract relevant implementation to perform.First of all, collect service instructions commanded to service robots in the average household, according to the Chinese grammar to remove obviously false and repeat, extracting the surface and deep information of the corpus collected, according to the natural language grammar to make up the learning corpus.Secondly, we analysis entity’s characteristic in the directive statements text, we use The Conditional Random Field Model to build entity recognition feature template, and then build entity recognition model, which is to complete entity identification of key information. Based on Support Vector Machine(SVM) model, using Lib SVM to build judge entity relationship model, which is to extract the entity of the relationship between the entities in the home environment, we extract key information and their relations, and store them in the relationship knowledge base, using this way to improve the efficiency of executing instructions.Finally, we extract key information and their relations, and store them in the relationship knowledge base, using this way to build the XML knowledge base. We show instructions recognition and the main steps in the process of building a knowledge base of experimental results and analysis, we combine with the specific example to show from text instructions to home service robot motion sequence recognition process. |
