mRAT-SQL+GAP: A Portuguese Text-to-SQL Transformer

The translation of natural language questions to SQL queries has attracted growing attention, in particular connection with transformers and similar models. A large number techniques are geared towards the English language; this work, we thus investigated when input given Portuguese language. To do so, properly adapted state-of-the-art tools resources. We changed RAT-SQL+GAP system by relying on a multilingual BART model (we report tests other models), produced translated version Spider dataset. Our experiments expose interesting phenomena that arise non-English languages targeted; particular, it is better train original training datasets together, even if single target desired. This fine-tuned double-size dataset (English Portuguese) achieved 83% baseline, making inferences for test investigation can help researchers produce results Machine Learning different from English. ready data available, open-sourced as mRAT-SQL+GAP at: https://github.com/C4AI/gap-text2sql.