Download link for the database

The train.csv and val.csv files contain the training and dev data for this assignment. Each query is associated with a db_id.

The details of the corresponding db_id are contained in the tables.json file. The table_names entry denotes the names of all tables. The column_names have an (table_index, column_name) tuple for each column which links them to the appropriate table. The foreign_keys entry indicates the column IDs which form a foreign key. Similarly the primary_keys entry denotes the primary key for each table. Rest of the entries are self-explanatory.

To evaluate your inferences, enter the evaluator directory and use the following command:

python evaluation.py --gold <GOLD_FILE> --pred <PRED_FILE> --db ../database/ --table ../tables.json --etype all

An example of a <GOLD_FILE> and <PRED_FILE> is given in the evaluator folder.