Research And Implementation Of Natural Language Instructions Parsing Based On Group Chat Information

In recent years,with the development of artificial intelligence technology and robotics,how to improve the intelligence of robots has become a hot topic of discussion in the industry.However,current robot intelligence solutions often focus only on sensors with sensing capabilities and related algorithms,ignoring the need to obtain information from sources such as instant messaging software in the mobile Internet.To address these problems,this paper relies on the National Key R&D Program of Human-Robot Intelligence Integration Technology to study the use of group chat implicit information by mobile robots to accomplish daily tasks,and validates the research by designing and building a simulation demonstration platform and a physical experiment platform.The main contributions of this paper are as follows:Firstly,we designed a framework for Chinese natural language instruction parsing task in order to realize the mapping from Chinese natural language instructions to robot actions.We constructed a Chinese natural language instruction dataset,and then proposed a BERTbased natural language instruction parsing scheme for the natural language parsing task.Finally,through model comparison experiments,we selected the optimal model for the metrics and verified the rationality and effectiveness of the proposed framework.Secondly,to make the robot possess the ability to obtain information from instant messaging software,we designed a task framework for group chat implicit information extracting.We construct a Chinese group chat conversation dataset,and then propose a group chat implicit information extraction solution for the constructed dataset to extract the implicit information(person,time,and location information)in the group chat.At the same time,the information presented in the group chat is fed into the dynamic spatio-temporal map proposed in this paper in real time,which solves the problem of how the robot can update its own knowledge base using the group chat information.Finally,we proved the rationality and effectiveness of the proposed dynamic spatio-temporal mapping through simulation experiments.Then,we designed and built a simulation demonstration platform and a physical experiment platform for Mobile Robot,and performed functional verification.First,we developed an intelligent mobile robot visualization platform based on Qt for demonstrating the effect of instruction parsing and group chat information extraction tasks.Next,we built a mobile robot physical experiment platform based on the requirements of the natural language instruction parsing task framework and the group chat implicit message extraction task framework,and connected the robot to the WeChat platform.We verified on the physical platform that the robot can extract information from group chats,update its own knowledge base in real time,and parse instructionsFinally,we concluded and summarized the whole paper and looked at the ways of improvement and the prospects of development of intelligent robots in obtaining information from mobile connectivity,taking into account the shortcomings of existing work and the trends of various tasks today.