Table structure recognition: a review of current approaches and future opportunities

Table Structure Recognition (TSR) refers to the process of detecting and understanding the internal layout of a table in a document. It can be divided into three subtasks, that is, table detection, table structure recognition and text extraction. There are a lot of methods for these tasks. Therefore, this paper systematically combines the existing literature on table structure recognition from the year 2021 to 2024. The aim of the study is to review the extant literature, identify the research gaps and pave the way for future study. To carry out this study, papers were extracted from the web of science and scopus database using keywords-table structure recognition and table detection. This review provides information about the various publically available datasets like ICDAR2013, TableBank, PubTabNet, etc. Furthermore, the study explores various recently developed models for the three subtasks of table recognition, along with their corresponding evaluation methods. The performance of several methods like ClusterNet, TableStrRec, TFLOP, etc. is also reported, highlighting their significance, strengths, and limitations. To sum up, this paper surveys existing dataset, evaluation strategies, recognition methods and explores potential future developments in the field of table recognition.