Relationship extraction model based on natural language reasoning

Abstract

The invention belongs to the technical field of automatic recognition, and particularly relates to a relation extraction model based on natural language reasoning, which comprises a Description Layer; an Encoder Layer; an Inferece Layer; and a Classification Layer. According to the model, a comparison test is carried out on a disclosed data set SemEval 2010Task-8 and four currently advanced models: 1) a GCN-based FAT-RE model, 2) a CNN and attention mechanism-based Att-Pooling-CNN model, 3) a BERT-based R-BERT model and 4) a BERT-based KnowBERT model, so that the models integrate information in a knowledge base, and the F1 score of the models reaches 90.1% and is higher than that of the other four models, the performance of the model is effectively improved by constructing relation description and multi-loss function superposition, priori knowledge is injected into the model, and in the reasoning process, key information in a target sentence is selected and noise in the target sentence is filtered according to the relation description.

Description

technical field [0001] The invention relates to the technical field of automatic identification, in particular to a relationship extraction model based on natural language reasoning. Background technique [0002] The Institute of Pattern Recognition Automation, Chinese Academy of Sciences proposed the PCNN model, which extracts the relationship between entities through convolutional neural networks and maximum pooling, and introduces remote supervision based on multi-instance learning to alleviate the problem of insufficient data. Due to the poor parallelism of traditional RNN, Google proposed a codec Transformer based entirely on the attention mechanism, which not only speeds up the operation speed, but also improves the accuracy of the model. Alibaba proposed the R-BERT model, which achieved good results by using two entities and the word vector represented by Bert's special mark CLS as a relational vector representation. Huawei's REDN model calculates the correlation bet...

AI Technical Summary

Problems solved by technology

Their performance is limited since the target sentences may not provide sufficient evidence and information

[0005]2) When injecting additional information into the model, existing methods often adopt pre-training methods, or inject information through natural language processing tools, which makes the model need more More training time and additional data sets, while the performance of the model will also depend on the performance of natural language processing tools

Images