| --- |
| license: apache-2.0 |
| dataset_info: |
| features: |
| - name: question |
| dtype: string |
| - name: table_info |
| dtype: string |
| - name: sql_query |
| dtype: string |
| splits: |
| - name: train |
| num_bytes: 847766 |
| num_examples: 3473 |
| download_size: 391731 |
| dataset_size: 847766 |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: data/train-* |
| --- |
| |
|
|
| # NSText2SQL Dataset (Reformatted for Fine Tuned Generative Models) |
| This is the exact same dataset as NSText2SQL: https://huggingface.co/datasets/NumbersStation/NSText2SQL, but with the data reformatted to allow direct use to fine tune generative models. The original license and credits for the original dataset remain in place. |
|
|
| Specifically, the changes from standard NSText2SQL are: |
|
|
| 1. Removed non-english questions |
|
|
| 2. Removed all rows with more than one input table, simplifying the problem for smaller models. |
|
|
| 3. Updated SQL queries in the dataset to prefer using LIKE statements for string matches, to allow better partial matching of results in chat scenarios where a user may not fully specify all data. |
|
|
| 4. Removed syntactically invalid SQL. Specifically, we created in-memory (SQLite) tables using the SQL DESCRIBE of the tables, then ran the SQL query against these in-memory tables. Any SQL queries that threw exceptions for any reason were discarded, and the rest that ran without exceptions were included in this dataset. |
|
|
