MEMORY-BASED REINFORCEMENT LEARNING METHOD FOR STORING OPTIONAL INFORMATION IN STREAMING DATA AND SYSTEM THEREFORE

摘要

A memory-based reinforcement learning method and system capable of selectively storing information in streaming data are disclosed. A question answering method using memory-based reinforcement learning according to an embodiment of the present invention includes: receiving streaming data for an input context input from a user in an episodic memory reader; analyzing the received streaming data in the episode memory reader and storing preset meaningful information used for question and answer in an external memory; and determining a response to the inputted query based on meaningful information stored in the external memory in a pre-learned question-and-answer model when a query is input from the user.