Commit
•
3793e64
1
Parent(s):
acfdebd
Create new file
Browse files
README.md
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
This is a reproduction of the CC-stories dataset as it has been removed from its original source.
|
| 2 |
+
To create this reproduction we process the English common crawl and only keep the top 0.1% of documents measured by their ngram overlap with a source document.
|
| 3 |
+
The source document is created by joining the queries from [PDP-60](https://cs.nyu.edu/~davise/papers/WinogradSchemas/PDPChallenge2016.xml) and [WSC273](https://cs.nyu.edu/~davise/papers/WinogradSchemas/WSCollection.xml). Note, as the original dataset does not mention removing duplicate queries, neither do we.
|
| 4 |
+
|
| 5 |
+
Following the filtering to have top documents we filter to only contain those and produce the dataset which features 2,105,303 lines and 153,176,685 words.
|