NLQA Based Knowledge Search for Virtual Geographic Environment

Abstract

Big data has injected new vitality into the virtual geographic environment (VGE). The intelligent virtual geographic environment system has put forward higher requirements for the interactivity and intelligent services. Firstly, in this paper, we have constructed a multi-level semantic conversion model, and enhanced the Chinese geographic knowledge graph by structured geographic information data such as the vector data of map. And then, based on Chinese geographic knowledge graph, we have proposed a bilateral LSTM-CRF model to achieve natural language question answering for VGE. Combing geographic knowledge base with virtual geographic scenes, we experimented the method. The experimental results prove that the method which is natural language question answering (NLQA) combined with the knowledge base can shorten the distance between people and virtual scenes and enhance the immersion and interactivity of VGE. It is an important way for virtual geographic environment to become intelligent.