Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,35 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
language:
|
4 |
+
- en
|
5 |
+
task_categories:
|
6 |
+
- summarization
|
7 |
+
- text2text-generation
|
8 |
+
tags:
|
9 |
+
- summarization
|
10 |
+
- long-document
|
11 |
+
pretty_name: SQuALITY v1.3
|
12 |
---
|
13 |
+
|
14 |
+
|
15 |
+
# SQuALITY - v1.3
|
16 |
+
|
17 |
+
> Original paper [here](https://arxiv.org/abs/2205.11465)
|
18 |
+
|
19 |
+
This is v1.3, the 'text' edition `.jsonl` files. See description from the [original repo](https://github.com/nyu-mll/SQuALITY):
|
20 |
+
|
21 |
+
> v1.3 fixes some bugs in v1.2. In v1.2, 10 out of 127 articles (each ~5k-word-long) are missing a few hundreds words each, so summaries may not be fully contained in the article. To fix this issue, we have updated the 10 articles.
|
22 |
+
|
23 |
+
## contents
|
24 |
+
|
25 |
+
> again, this is taken from the repo
|
26 |
+
|
27 |
+
Each data file ({train/dev/test}.jsonl) is formatted as a JSON lines file. Each row in the data file is a JSON dictionary with the following fields:
|
28 |
+
|
29 |
+
- metadata: the Gutenberg story ID, an internal UID, and the Project Gutenberg license
|
30 |
+
- document: the Gutenberg story
|
31 |
+
questions: a list of questions and accompanying responses
|
32 |
+
- question text
|
33 |
+
- question number: the order in which that question was answered by the writers
|
34 |
+
- responses: list of worker's response, where each response is a dictionary containing the (anonymized) worker ID, an internal UID, and their response to the question
|
35 |
+
|