How far can we get by training Text-to-SQL models without any annotated SQL? Pretty far, it seems!
Our new work in Findings of #NAACL2022 with @JonathanBerant and Daniel Deutch
Paper: arxiv.org/abs/2112.06311
Code: github.com/tomerwolgithub…
🧵 1/5
Our new work in Findings of #NAACL2022 with @JonathanBerant and Daniel Deutch
Paper: arxiv.org/abs/2112.06311
Code: github.com/tomerwolgithub…
🧵 1/5
Text-to-SQL models should help non-experts easily query databases. But annotating examples to train them requires expertise (labeling NL utterances with SQL queries).
Can we train good enough models without any expert annotations?
2/5
Can we train good enough models without any expert annotations?
2/5
Instead of gold SQL, we train text-to-SQL models on weak supervision: (1) answers & (2) question decompositions (annotated / predicted by a model) ⛏️
3/5
3/5
Using the database + question decomposition + answer we automatically synthesize a corresponding SQL query. This is well captured by mapping rules from the decomposition to SQL.
4/5
4/5
We test on five text-to-SQL benchmarks:
(1) Weakly supervised models reach ~94% of those trained on gold SQL
(2) Even models trained on few / zero in-domain decompositions still reach ~90% of the gold SQL ones
More results in our paper! 📃⚒️
👉🏽arxiv.org/pdf/2112.06311…
5/5
(1) Weakly supervised models reach ~94% of those trained on gold SQL
(2) Even models trained on few / zero in-domain decompositions still reach ~90% of the gold SQL ones
More results in our paper! 📃⚒️
👉🏽arxiv.org/pdf/2112.06311…
5/5
• • •
Missing some Tweet in this thread? You can try to
force a refresh
