An Interactive NL2SQL Approach With Reuse Strategy (The More You Ask, The Better You Query)

Data

Downlaod data from SParC. And then put in sparc/ folder.

• What is SParC? SParC is a dataset for cross-domain Semantic Parsing in Context. It is the context-dependent/multi-turn version of the Spider task, a complex and cross-domain text-to-SQL challenge. SParC consists of 4,298 coherent question sequences (12k+ unique individual questions annotated with SQL queries annotated by 14 Yale students), obtained from user interactions with 200 complex databases over 138 domains.

Preporcessing data

Running the follow scripts, and then you will get preprocessed pre_train.json and pre_dev.json.

sh run_preprocess.sh path/to/save

Train

Running the follow command:

python3 train.py --cuda --epoch 21 \
--loss_epoch_threshold 10 \
--sketch_loss_coefficient 0.5 \
--column_pointer --use_query_attention --fine_tune_bert\
--action_embed_size 128 \
--type_embed_size 128 \
--maximum_utterances 6 \
--use_schema_encoder --use_copy_switch\
--encoder_state_size 300 \
--hidden_size 300 \
--encoder_num_layers 1 \