Network table structure identification method and device, computer device and computer readable storage medium

Abstract

The invention discloses a network table structure identification method and device, a computer device and a computer readable storage medium, and the method comprises the steps: preprocessing an input HTML file to obtain table related information in the HTML file, the table related information comprising a cell text and the position of a cell, then, using the trained network table structure recognition model for recognizing the network table structure on the basis of the obtained cell text and the position where the cell is located, and since the cell is adopted as the minimum recognition granularity, compared with an existing mode that a table line is adopted as a recognition unit, the accuracy of a recognition result is greatly improved, and the recognition efficiency is improved. And the method can adapt to various complex network table structure identification scenes.

Description

technical field [0001] The present invention relates to the technical field of form information extraction, in particular to a method and device for identifying a network form structure, a computer device and a computer-readable storage medium. Background technique [0002] As an important form of information representation, web tables widely exist in web documents, which store a large amount of high-value information. However, due to the lack of clear semantic information and complex and diverse table structures, it is difficult for computers to accurately analyze the table content. understand. Therefore, the research on information extraction from web forms is of great significance. The table structure recognition is one of the research hotspots in the field of table information extraction. It specifically refers to the analysis of the table structure, the division of the table area, etc., such as including header area identification, table body area identification, etc. ...

AI Technical Summary

Problems solved by technology

[0005] However, most of the existing table structure recognition algorithms are aimed at simple tables in a specific field, and the existing recognition algorithms usually identify the table structure in units of table rows, but the network tables in real scenarios are extremely complex, and within a table row Table header cells and table body cells may appear at the same time. Simply dividing the table structure for table rows cannot meet the needs of real scenarios, resulting in low accuracy of existing table structure recognition algorithms.

At the same time, the existing table structure recognition algorithm can only be applied to specified fields, and it is difficult to migrate to other fields. Among them, when the rule-based table extraction algorithm is migrated to other fields, it is necessary to ask experts in the field to re-specify the rules. Algorithm development The cycle is long and the accuracy of the algorithm is generally not high, and when the table extraction algorithm based on machine learning is migrated to other fields, the data needs to be re-labeled, which requires a lot of manpower and time costs

Images