Table-To-Text generation and pre-training with TabT5
This addresses the problem of generating text from tables for applications like QA and data-to-text, representing a novel method rather than an incremental improvement.
The paper tackles the limitation of encoder-only transformer models in table understanding tasks by introducing TABT5, an encoder-decoder model that generates natural language text from tables, achieving state-of-the-art results with improvements such as a 15% increase in sequence accuracy for spreadsheet formula prediction.
Encoder-only transformer models have been successfully applied to different table understanding tasks, as in TAPAS (Herzig et al., 2020). A major limitation of these architectures is that they are constrained to classification-like tasks such as cell selection or entailment detection. We present TABT5, an encoder-decoder model that generates natural language text based on tables and textual inputs. TABT5 overcomes the encoder-only limitation by incorporating a decoder component and leverages the input structure with table specific embeddings and pre-training. TABT5 achieves new state-of-the-art results on several domains, including spreadsheet formula prediction with a 15% increase in sequence accuracy, QA with a 2.5% increase in sequence accuracy and data-to-text generation with a 2.5% increase in BLEU.