Data Augmentation on Graphs for Table Type Classification

Tables are widely used in documents because of their compact and structured representation information. In particular, scientific papers, tables can sum up novel discoveries summarize experimental results, making the research comparable easily understandable by scholars. Since layout is highly variable, it would be useful to interpret content classify them into categories. This could helpful directly extract information from for instance comparing performance some models given paper result tables. this work, we address classification using a Graph Neural Network, exploiting table structure message passing algorithm use. We evaluate our model on subset Tab2Know dataset. contains few examples manually annotated, propose data augmentation techniques graph structures. achieve promising preliminary proposing method suitable graph-based representation.