Reinforcement learning knowledge graph reasoning method based on action sampling

Abstract

The invention discloses a reinforcement learning knowledge graph reasoning method based on action sampling. Aiming at the problems of insufficient representation capability, invalid redundant action selection and no memory component in the traditional knowledge graph reinforcement learning reasoning algorithm, the method comprises the following steps of: predicting a score according to an original fact of a representation learning method on a data set; a representation learning method with higher adaptability is selected in a targeted manner to represent a reinforcement learning environment so as to enhance the algorithm representation capability; designing an action sampler to reduce invalid redundant action selection of the intelligent agent in the migration process; the LSTM is used as a memory component, and historical information is coded to increase the model precision, so that the algorithm can obtain an effect superior to that of a path-based reasoning algorithm under the condition of getting rid of pre-training. According to the method, a path obtained by walking of an intelligent agent in an environment is mapped to a three-layer LSTM strategy network, the intelligent agent is promoted to select a more meaningful path through action sampling, and finally relatively accurate entity relationship path learning is realized.

Description

technical field [0001] The invention belongs to the field of natural language processing. Background technique [0002] In recent years, deep learning techniques have achieved many state-of-the-art results on various classification and recognition problems. However, complex natural language processing problems often require multiple interrelated decisions, making deep learning models capable of learning reasoning remains a challenging problem. To handle complex queries without obvious answers, intelligent machines must be able to reason with existing resources and learn to infer unknown answers. [0003] With the continuous development of knowledge graph reasoning technology, reinforcement learning has been proved to achieve better results in knowledge reasoning tasks. DeepPath released by EMNLP2017 introduced reinforcement learning into the reasoning of knowledge graph for the first time. It simply sampled the knowledge graph and put it into the policy network for trainin...

AI Technical Summary

Problems solved by technology

[0008] The present invention proposes an action sampling-based reinforcement learning knowledge map reasoning method, aiming to solve the problems of insufficient representation ability, invalid action selection, and no memory components of existing reinforcement learning reasoning methods

Images