TABBIE: Pretrained Representations of Tabular Data
This work addresses the need for efficient and effective tabular representation learning for tasks like missing cell population, offering a domain-specific improvement over prior methods.
The paper tackled the problem of pretraining representations for tabular data, showing that existing joint text-table methods underperform on tasks without associated text, and introduced TABBIE with a corrupt cell detection objective that achieves state-of-the-art results on table-based prediction tasks.
Existing work on tabular representation learning jointly models tables and associated text using self-supervised objective functions derived from pretrained language models such as BERT. While this joint pretraining improves tasks involving paired tables and text (e.g., answering questions about tables), we show that it underperforms on tasks that operate over tables without any associated text (e.g., populating missing cells). We devise a simple pretraining objective (corrupt cell detection) that learns exclusively from tabular data and reaches the state-of-the-art on a suite of table based prediction tasks. Unlike competing approaches, our model (TABBIE) provides embeddings of all table substructures (cells, rows, and columns), and it also requires far less compute to train. A qualitative analysis of our model's learned cell, column, and row representations shows that it understands complex table semantics and numerical trends.