Speech-to-SQL Parsing on MASpider Custom Data generalization (test)
67.8SELECT Component AccuracyS2SQL-TTS
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| S2SQL-TTSSynthesized data upper bound=FastSpeech 22023.05 | 67.8 | 45.5 | 53.6 | 69.2 | 96.6 | 73.1 | 35.7 | |
| CascadedASR=wav2vec 2.0, Text-to-SQL parser=RAT-SQL2023.05 | 61.9 | 45.2 | 50.5 | 61.7 | 96.2 | 73.3 | 30.4 | |
| Wav2SQLSpeech feature extractor=Hubert, Language model=GloVe2023.05 | 61.1 | 38.6 | 51.5 | 62.3 | 95.8 | 63.1 | 28.6 | |
| DeepSpeechSQLSpeech encoder=DeepSpeech2023.05 | 41 | 24.2 | 25.5 | 18.1 | 95.1 | 31.8 | 12.2 |