Text-to-Text Pre-Training for Data-to-Text Tasks
This provides a useful baseline for data-to-text tasks, benefiting researchers in natural language generation, but it is incremental as it applies an existing pre-training method to a specific domain.
The paper tackles the problem of data-to-text generation by applying text-to-text pre-training with T5, showing that it outperforms specialized pipelined architectures and other pre-training methods like BERT and GPT-2, with large improvements on out-of-domain test sets.
We study the pre-train + fine-tune strategy for data-to-text tasks. Our experiments indicate that text-to-text pre-training in the form of T5, enables simple, end-to-end transformer based models to outperform pipelined neural architectures tailored for data-to-text generation, as well as alternative language model based pre-training techniques such as BERT and GPT-2. Importantly, T5 pre-training leads to better generalization, as evidenced by large improvements on out-of-domain test sets. We hope our work serves as a useful baseline for future research, as transfer learning becomes ever more prevalent for data-to-text tasks.