Reasoning system, method and device based on multi-strategy intelligent agent and electronic equipment

Abstract

The invention provides a reasoning system, method and device based on a multi-strategy intelligent agent and electronic equipment. The method comprises the steps of S101, encoding a local graph through a DBL-LSTM encoder and outputting state encoding information; S102, performing action selection from a relation set of entities in the search window according to the state coding information through an MLP classifier to obtain an expanded new node; S103, updating a local graph and state coding information according to the new node; S104, feeding back a reward obtained by action selection to the multi-strategy agent, and accumulating the reward to a reward function; S105, executing the steps S102 to S104 are executed in parallel, and the reward function is maximized through the REINFORCE algorithm; S106, respectively generating a first local inference graph, a second local inference graph and a third local inference graph of a given head entity and relationship; S107, performing graph combination to obtain a final local graph; and S108, determining a tail entity according to the final local graph. The reasoning method solves the problems that in the prior art, after an invalid path is selected, selection cannot be stopped again, and all possible paths cannot be noticed at the same time.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a reasoning system, method, device and electronic equipment based on a multi-strategy agent. Background technique [0002] The knowledge graph is essentially a semantic network. Due to the large amount of high-quality knowledge, the knowledge graph plays a vital role in various industries, such as recommendation systems, intelligent question answering, etc. Since the knowledge graph provides both explicit and implicit information, even though both head entities and tail entities are included in the knowledge graph, there is no direct relationship between them. While hidden relationships can be obtained based on reasoning, machines lack this ability. [0003] Phase-based path reasoning mainly faces two challenges. First, when the agent chooses an invalid path during its reasoning, it does not stop and choose again, but repeats the same mistake. The reason is tha...

AI Technical Summary

Problems solved by technology

[0004] The object of the present invention is to provide a reasoning system, method, device and electronic equipment based on a multi-strategy agent. choice, and the problem of not being able to notice all possible paths simultaneously

Images