Table information extraction model training method and device

Abstract

The invention discloses a table information extraction model training method, which comprises the following steps: processing table cells of table corpus to obtain feature vectors of the table cells; calculating an adjacent matrix according to the position information of the cells, and performing feature extraction on the feature vectors of the cells and the adjacent matrix to obtain high-order feature vectors of the cells; predicting an original text of the cell by using the high-order feature vector, and performing model training by using the text of the cell to obtain a table language model; and training the table language model by using the training sample corresponding to the current table information extraction task to obtain a table information extraction model. For each different table extraction task, only a small number of training samples corresponding to different tasks need to be used for training to obtain the corresponding table information extraction model on the basis of the trained table language model, and the table language model does not need to be trained in each training, so that the training time and the training samples are obviously reduced.

Description

technical field [0001] This application relates to the technical field of artificial intelligence, in particular to a training method and device for a form information extraction model. Background technique [0002] With the increasing popularity of information technology, the demand for extracting information from tables has become increasingly prominent. Currently, rule algorithms are mainly used to extract information from tables. However, due to the lack of generalization of the rule algorithm, the corresponding rule algorithm needs to be redeveloped for each type of table and different extraction requirements, which not only makes the development task heavy, the algorithm reuse value is low, but also leads to different extraction task adaptability Poor, the extraction efficiency of table information is low. Contents of the invention [0003] The present application provides a training method and device for a form information extraction model to solve one or more tec...

AI Technical Summary

Problems solved by technology

However, due to the lack of generalization of the rule algorithm, the corresponding rule algorithm needs to be redeveloped for each type of table and different extraction requirements, which not only makes the development task heavy, the algorithm reuse value is low, but also leads to different extraction task adaptability Poor, low extraction efficiency of table information

Images