Method and device for training information extraction model and obtaining knowledge graph, and equipment for training information extraction model and obtaining knowledge graph

Abstract

The invention relates to the technical field of computers, and provides a method and device for training an information extraction model and obtaining a knowledge graph and equipment for training theinformation extraction model and obtaining the knowledge graph, which are used for improving the robustness of the information extraction model. The method comprises: training an initial first information extraction model, and updating the first information extraction model based on a prediction result of a noise sample to obtain a first intermediate state model; updating the first information extraction model based on the difference between the prediction result of the noise sample by the first intermediate state model and the prediction result of the non-noise sample by the initial second information extraction model to obtain a second intermediate state model; updating the second intermediate state model based on the prediction result of the non-noise sample by the second intermediate state model to obtain a reference model; and adjusting parameters of the reference model based on a preset smooth coefficient to obtain a target information extraction model. The model parameters are updated based on the meta-learning mode, the updated model is more robust, and the constructed knowledge graph is more accurate.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to the field of machine learning technology, and provides a method, device and equipment for training information extraction models and acquiring knowledge graphs. Background technique [0002] With the rapid development of the Internet today, it is an era dominated by information and big data. How to extract the content of concern in the era of information explosion is very important. The research on information extraction aims to provide people with more powerful information acquisition tools to deal with the information explosion. serious challenge to come. At present, the most commonly used method is the remote supervision relation extraction method, which maps the existing knowledge base to rich unstructured data, thereby generating a large amount of training data, and then training the information extraction model. However, distant supervision inevitably suffers f...

AI Technical Summary

Problems solved by technology

However, remote supervision inevitably has the problem of data labeling errors, generating noisy data with wrong labels

Because neural networks are prone to overfitting to noisy data, training an information extraction model on these noisy data can seriously affect the performance of the model

[0003] In summary, the current information extraction model is less robust to noisy data, and it is easy to fit noisy data, resulting in performance loss of the information extraction model

Images