Method for optimizing data understanding and processing by utilizing semantic association among multiple models

Abstract

The invention discloses a method for optimizing data understanding processing by utilizing semantic association among multiple models, which comprises the following steps of: configuring multiple dataunderstanding models, and selecting multiple segments of original data; using all configured data understanding models for marking each section of original data, and obtaining a processing result ofeach data understanding model for each section of original data to serve as multi-section training data; training a dynamic Bayesian network and a reinforcement learning decision network by using themultiple segments of training data to obtain a model semantic association network; and forming a multi-model data understanding processing optimization control rule by using the model semantic association network, and selecting an optimal data understanding model for data understanding processing according to input data by using the multi-model data understanding processing optimization control rule. According to the method, semantic association between the models is used for carrying out optimal selection of understanding models on each section of unknown content data, and semantic recognition and understanding on the data are achieved as perfect as possible with the minimum model operation expenditure.

Description

technical field [0001] The invention relates to the field of data mining, in particular to a method for optimizing data understanding and processing by utilizing semantic association among multiple models. Background technique [0002] At present, a large amount of unlabeled and unknown raw data needs to be identified and labeled as perfect as possible by using machine learning. However, if the raw data of each piece of unknown content is identified one by one using various machine learning models, it will bring huge computing overhead and unnecessary waste of computing resources. In this regard, existing methods for optimizing data understanding and processing basically start from the following aspects: optimizing model structure, compressing model size, and building multi-task models. However, these methods all focus on the optimization of a single model, while ignoring the problem of multiple models that usually need to be used in large-scale data understanding and proce...

AI Technical Summary

Problems solved by technology

However, these methods all focus on the optimization of a single model, while ignoring the problem of multiple models that usually need to be used in large-scale data understanding and processing for data content identification. However, how to optimize data identification and understanding based on multiple models has not yet been related to technology.

Images