An Improved Method for Entity Linking in Simple Question Answering Based on Knowledge Graph

Abstract

The invention discloses an improved method for entity linking in simple questions and answers based on knowledge graphs, which belongs to the technical field of natural language processing. A central server and a question input client are established, and an entity detection module, an entity candidate set module, and a knowledge The map retrieval module and the entity matching module detect the problem data, establish an entity candidate set, encode the problem data, perform three-level coding on the entities in the entity candidate set, and select the entity candidate set with the highest matching score with the problem data. n entities, a unique way of encoding questions is adopted. The present invention proposes a method for encoding entities at three levels, which makes full use of the type information and name information of entities, combined with the way of encoding questions, can effectively Solve entity confusion and OOV issues.

Description

Technical field [0001] The present invention belongs to the field of large data technology, which relates to an improved method of entity link based on a simple question and answer of knowledge map. Background technique [0002] In recent years, there have been more and more open source knowledge maps (KG) containing a lot of facts, such as FreeBase, YAGO, and DBPEDIA. Questions and answers (KG-QA) in the source of knowledge maps are hot spots in recent years. There are two main methods for the storage of knowledge maps: based on RDF storage, based on the storage of the graph database. [0003] Traditional KG-QA methods can be divided into three Class KG-QA. The first category is a semantic parsing: the method is a method of biased LINGUISTIC. The main thinking is to convert the natural language into a series of formalized logical forms. By analyzing the logical form, it can express the entire The logical form of the problem is in the knowledge base in the knowledge base. The sec...

AI Technical Summary

Problems solved by technology

[0007] In recent years, some neural network models combined with attention mechanisms have been continuously proposed. In entity linking, the main task of this model is to make the vectorization of the problem better represent the information related to the entity, which is the information about the entity in the problem. Some can be utilized to maximize, but this model is generally more complex and does not deal well with the problem of entity confusion

Images