Datasets:
File size: 3,678 Bytes
890c458 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 | ---
license: cc-by-sa-4.0
task_categories:
- text2text-generation
- table-question-answering
language:
- en
- sql
tags:
- text-to-sql
- sql
- spider
- flan-t5
- seq2seq
- nlp
size_categories:
- 1K<n<10K
---
# SPIDER Text-to-SQL — Easy Access Version
A clean, HuggingFace-native version of the [SPIDER](https://yale-seas.yale.edu/spider/) Text-to-SQL benchmark. The original SPIDER dataset requires manually downloading a ZIP file from the Spider website. This version makes it instantly accessible via `load_dataset`.
## What's Included
Each row contains the question, gold SQL, the database identifier, and a pre-parsed compact schema string — everything needed to train or evaluate a Text-to-SQL model without any additional preprocessing.
| Column | Description |
|---|---|
| `db_id` | Database identifier (e.g. `"concert_singer"`) |
| `question` | Natural language question |
| `query` | Gold standard SQL answer |
| `db_schema` | Compact schema: `"table: col (type), col (type) | table2: ..."` |
| `question_toks` | Tokenized question words (list of strings) |
## Splits
| Split | Source file | Examples |
|---|---|---|
| train | `train_spider.json` | 7,000 |
| test | `train_others.json` | 1,034 |
> **Note**: Following standard SPIDER practice, `train_others.json` is used as the held-out evaluation set. The original SPIDER test set is withheld for the official leaderboard.
## Usage
```python
from datasets import load_dataset
dataset = load_dataset("YOUR_USERNAME/spider-text2sql")
train = dataset["train"]
test = dataset["test"]
# Access fields
example = train[0]
print(example["question"]) # "How many heads of the departments are older than 56?"
print(example["query"]) # "SELECT count(*) FROM head WHERE age > 56"
print(example["db_id"]) # "department_management"
print(example["db_schema"]) # "department: Department_ID (number), ... | head: ..."
```
## Schema Format
The `db_schema` column uses a compact linear format widely used in the Text-to-SQL literature:
```
table1: col1 (type), col2 (type), col3 (type) | table2: col4 (type), col5 (type)
```
This format is:
- Human-readable and model-friendly
- Fits within typical 512-token input limits for most seq2seq models
- Derived directly from the official SPIDER `tables.json`
## Fine-tuning Example (Flan-T5 prompt format)
This dataset pairs naturally with prompt-based fine-tuning:
```python
def build_prompt(example):
return (
f"Translate to SQL: {example['question']}\n"
f"Database schema:\n{example['db_schema']}"
)
# example["query"] is the target output
```
## Difference from Original SPIDER
| | Original SPIDER | This Dataset |
|---|---|---|
| Download method | Manual ZIP from website | `load_dataset(...)` ✅ |
| Schema included | Separate `tables.json` | ✅ Pre-joined per example |
| Complex `sql` dict | ✅ Included | ❌ Omitted (noisy for most use cases) |
| `query_toks_no_value` | ✅ Included | ❌ Omitted |
| Ready to train | Requires preprocessing | ✅ Yes |
## Source & License
- Original dataset: [SPIDER (Yu et al., 2018)](https://yale-seas.yale.edu/spider/)
- License: **Creative Commons Attribution-ShareAlike 4.0 (CC BY-SA 4.0)**
- This derived dataset is released under the same license.
## Citation
```bibtex
@inproceedings{yu-etal-2018-spider,
title = "{S}pider: A Large-Scale Human-Labeled Dataset for Complex and Cross-Domain Semantic Parsing and Text-to-{SQL} Task",
author = "Yu, Tao and others",
booktitle = "Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing",
year = "2018",
url = "https://aclanthology.org/D18-1425",
}
```
|