Datasets:
xhluca
commited on
Commit
·
af08558
1
Parent(s):
2757638
Update readme
Browse files
README.md
CHANGED
|
@@ -30,4 +30,38 @@ tags:
|
|
| 30 |
- convAI
|
| 31 |
---
|
| 32 |
|
| 33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
- convAI
|
| 31 |
---
|
| 32 |
|
| 33 |
+
|
| 34 |
+
# WebLINX: Real-World Website Navigation with Multi-Turn Dialogue
|
| 35 |
+
|
| 36 |
+
*Xing Han Lù, Zdeněk Kasner, Siva Reddy*
|
| 37 |
+
|
| 38 |
+
[Link to the website](https://mcgill-nlp.github.io/weblinx)
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
## Quickstart
|
| 42 |
+
|
| 43 |
+
To get started, simply install `datasets` with `pip install datasets` and load the chat data splits:
|
| 44 |
+
|
| 45 |
+
```python
|
| 46 |
+
from datasets import load_dataset
|
| 47 |
+
|
| 48 |
+
# Load the training, validation and test (IID) splits
|
| 49 |
+
train = load_dataset("McGill-NLP/weblinx", "train")
|
| 50 |
+
valid = load_dataset("McGill-NLP/weblinx", "valid")
|
| 51 |
+
test = load_dataset("McGill-NLP/weblinx", "test")
|
| 52 |
+
|
| 53 |
+
# Load one of the 4 out-of-domain splits (test_web, test_vis, test_geo, test_cat)
|
| 54 |
+
test_web = load_dataset("McGill-NLP/weblinx", "test_web")
|
| 55 |
+
```
|
| 56 |
+
|
| 57 |
+
## Raw Data
|
| 58 |
+
|
| 59 |
+
To use the raw data, you will need to use the `huggingface_hub`:
|
| 60 |
+
|
| 61 |
+
```python
|
| 62 |
+
from huggingface_hub import snapshot_download
|
| 63 |
+
|
| 64 |
+
snapshot_download(repo_id="McGill-NLP/WebLINX-full", repo_type="dataset", local_dir="./data/weblinx")
|
| 65 |
+
```
|
| 66 |
+
|
| 67 |
+
For more information on how to use this data using our [official library](https://github.com/McGill-NLP/WebLINX), please refer to the [WebLINX documentation](https://mcgill-nlp.github.io/WebLINX/docs).
|