ZhuofengLi commited on
Commit
0d3d276
·
verified ·
1 Parent(s): 87c05c3

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +31 -0
README.md ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ## Dataset
2
+ - split_survey_df: The split version of SciReviewGen, which aims to generate literature review **chapters**
3
+ - original_survey_df: The original version of SciReviewGen, which aims to generate **the entire text** of literature reviews
4
+ - summarization_csv: CSV files suitable for summarization task.
5
+
6
+ ### Data format
7
+ #### split_survey_df & original_survey_df
8
+ - Row:
9
+ - literature review chapter or the entire text of literature review
10
+ - Column:
11
+ - paper_id: paper_id used in [S2ORC](https://github.com/allenai/s2orc)
12
+ - title: title of the literature review
13
+ - abstract: abstract of the literature review
14
+ - section: chapter title
15
+ - text: body text of literature review chapter or literature review paper
16
+ - n_bibs: number of the cited papers that can be used as inputs
17
+ - n_nonbibs: number of the cited papers that cannot be used as inputs
18
+ - bib_titles: titles of the cited papers
19
+ - bib_abstracts: abstracts of the cited papers
20
+ - bib_citing_sentences: citing sentences that cite the cited papers
21
+ - split: train/val/test split
22
+
23
+ #### summarization_csv
24
+ - Row:
25
+ - literature review chapter
26
+ - Column:
27
+ - reference: `literature review title <s> chapter title <s> abstract of cited paper 1 <s> BIB001 </s> literature review title <s> chapter title <s> abstract of cited paper 2 <s> BIB002 </s> ...`
28
+ - target: literature review chapter
29
+
30
+ ### Data construction
31
+ Use `make_summarization_csv.py` to convert df to csv