Update README.md
Browse files
README.md
CHANGED
@@ -3,4 +3,38 @@ task_categories:
|
|
3 |
- summarization
|
4 |
- text-generation
|
5 |
- text2text-generation
|
6 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
- summarization
|
4 |
- text-generation
|
5 |
- text2text-generation
|
6 |
+
---
|
7 |
+
|
8 |
+
# DivSumm summarization dataset
|
9 |
+
|
10 |
+
Dataset introduced in the paper: Analyzing the Dialect Diversity in Multi-document Summaries (COLING 2022)
|
11 |
+
_Olubusayo Olabisi, Aaron Hudson, Antonie Jetter, Ameeta Agrawal_
|
12 |
+
|
13 |
+
|
14 |
+
DivSumm is a novel dataset consisting of dialect-diverse tweets and human-written extractive and abstractive summaries. It consists of 90 tweets each on 25 topics in multiple English dialects (African-American, Hispanic and White), and two reference summaries per input.
|
15 |
+
|
16 |
+
|
17 |
+
|
18 |
+
## Directories
|
19 |
+
|
20 |
+
input_docs - 90 tweets per topic evenly distributed among 3 dialects; total 25 topics
|
21 |
+
|
22 |
+
abstractive - Two annotators were asked to summarize each topic in 5 sentences using their own words.
|
23 |
+
|
24 |
+
extractive - Two annotators were asked to select 5 tweets from each topic that summarized the input tweets.
|
25 |
+
|
26 |
+
|
27 |
+
## Paper
|
28 |
+
|
29 |
+
|
30 |
+
You can find our paper [here](https://aclanthology.org/2022.coling-1.542/). If you use this dataset in your work, please cite our paper:
|
31 |
+
|
32 |
+
|
33 |
+
@inproceedings{olabisi-etal-2022-analyzing,
|
34 |
+
title = "Analyzing the Dialect Diversity in Multi-document Summaries",
|
35 |
+
author = "Olabisi, Olubusayo and Hudson, Aaron and Jetter, Antonie and Agrawal, Ameeta",
|
36 |
+
booktitle = "Proceedings of the 29th International Conference on Computational Linguistics",
|
37 |
+
month = oct,
|
38 |
+
year = "2022",
|
39 |
+
}
|
40 |
+
|