| With the rapid development of science and technology,intelligent robots for family services become members of the family.Although they provide a great deal of help for older and disabled members of the family in our daily life,there are still many key issues have not been resolved properly.In this paper,we studied the Chinese instructions and the key technology of complex instruction analysis of home service robot,to give a complete technology solutions which was used to built a complex robot action sequence of instructions.Firstly,the corpus must be separate and mark in pretreatment work according to the characteristics of the family service instructions.In order to extract command key information,we combined the conditional random field(CRF)and rules.Regular expressions and CRF statistical model were used to construction entity recognition feature templates of family service instructions.Then we can identify the key information of instructions when the target entity,service entity,object entity,and qualified entity were identified.Secondly,after the complex instructions was identified by the key information,there will be a large number of qualified entities(qualified information which was required to complete a command object).The knowledge base of resource description framework(RDF)was built to limit the processing entities.The data of family service was stored in the triple pattern.On the basis of results recognized by named entity,we have the property term disambiguation and form the instruction semantic map.The query algorithm auto-generated by SPARQL(Simple Protocol and RDF Query Language)allow robot do some query operation in the knowledge base,so that the service robot possess the ability of reasoning depend on qualifying information.,choosing the correct object,and returning a corresponding sequence of instruction execution.Finally,we display our research and the results of the key experiments systematically.The results verified the practicality of our method. |
PDF Full Text Request