RuTaBERT is a model solving the problem of Column Type Annotation with pre-trained large language model (BERT), trained on the Russian corpus.

Dataset from the authors of RuTaBERT and is based on the Russian Web Tables. Only relational tables were chosen from Russian Web Tables with headers matching their selected 170 DBpedia semantic types.

CoLeM framework is a table model based on contrastive learning techniques for solving the problem of Column Type Annotation.