Table 2 Evaluation results of the first experiment
From: Comparative study on the customization of natural language interfaces to databases
NLIDB | Total queries | Correct w/initial customization | Correct w/fine-tuning | Minutes for customizing one query | Recall w/initial customization (%) | Recall w/fine-tuning (%) |
|---|---|---|---|---|---|---|
ELF | 70 | 7 | 8.28 | 93.4 | 10 | 11.83 |
Our NLIDB | 70 | 17 | 31.28 | 8.4 | 24.28 | 44.69 |