TD4CLTabs Corpus

TD4CLTabs (Table Detection for Computational Linguistics Tables) is a novel domain-specific table type classification dataset, which contains 13,000 table images (10,000 for training and 3,000 for testing) and their corresponding type labels obtained from ACL Anthology. The dataset was constructed using the novel taxonomies which are based on the layout and structure characteristics of tables found in scholarly publications. Detailed information can be found in the following paper: He, J., Borisova, E., & Rehm, G. 2024. Towards a Novel Classification of Table Types in Scholarly Publications. Accepted to the Natural Scientific Language Processing and Research Knowledge Graphs Workshop (NSLP 2024). The folder train, test, and SciTSRComp contain the table images and folder metadata contains the label information for each taxonomy on each dataset. In metadata, labels_metadata.json stores all the taxonomy mappings.