Family-Oriented Service Robots Chinese Instruction Depth Information Recognition System

Intelligent service robot is gradually from the screen to our life, which has brought us convenience, efficiency. At the same time, there are still many problems to be solved. This paper studied deeply on the service robots Chinese instruction deep information recognition.First of all, in order to guide the robot to complete the deep service instruction purposes through natural language, this paper proposes a combined use of named entity recognition and the knowledge base of deep information identification. First in-depth study of the collected family services instruction natural language corpus, to analyze the characteristics of each instruction entity text. On the basis of using CRFs, construct entity recognition feature template against family services instruction, build entity recognition model and finish surface information identification. In order to solve the problem of deep information identification, the paper introduces the family service instruction knowledge base, from the relationship between matter and matter, matter and instruction of the family environment, structured the deep information of service instruction. Through the instruction knowledge base established the corresponding relation by the surface information to the deep information, so as to obtain deep instruction information and construct the instruction execution sequence.Secondly, for learning unknown instruction, this paper puts forward a kind of method of learning unknown instruction based on automatic entity relation extraction.To endow robots with the ability to learn unknown instruction. Through in-depth study of the instruction learning corpus, the problem is transformed to entity relation extraction. Based on the theory of support vector machine, we use Lib SVM to build entity relationship determine model in the area of family services, and use the model to achieve the work of entity relation extraction and complete the unknown instruction learning task.Finally, we show the experiment results and analysis of the main links of instruction recognition and the learning process and combine the concrete examples to demonstrate the identification process from the voice commands to the robot motion sequences.