PaulD commited on
Commit
17e101a
1 Parent(s): 2dc11ed

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +3 -3
README.md CHANGED
@@ -2,14 +2,14 @@
2
  license: mpl-2.0
3
  widget:
4
  - text: "The closest planet to earth is <mask>."
5
- - text: "Electrical power is generated on a spacecraft with <mask>."
6
  ---
7
 
8
  ### CosmicRoBERTa
9
 
10
- This model is a further pre-trained version of RoBERTa on a domain-specific corpus, including abstracts from the NTRS library, abstracts from SCOPUS etc...
11
 
12
- The further pre-training corpus includes publications abstracts, books, and Wikipedia pages related to space systems. Corpus size is 14.3 GB. SpaceRoBERTa was further pre-trained on this domain-specific corpus from [RoBERTa-Base](https://huggingface.co/roberta-base). In our paper, it is then fine-tuned for a Concept Recognition task.
13
 
14
  ### BibTeX entry and citation info
15
 
 
2
  license: mpl-2.0
3
  widget:
4
  - text: "The closest planet to earth is <mask>."
5
+ - text: "Electrical power is stored on a spacecraft with <mask>."
6
  ---
7
 
8
  ### CosmicRoBERTa
9
 
10
+ This model is a further pre-trained version of RoBERTa for space science on a domain-specific corpus, which includes abstracts from the NTRS library, abstracts from SCOPUS, ECSS requirements, and other sources from this domain.
11
 
12
+ The model performs slightly better on a subset of the CR task presented in our paper.
13
 
14
  ### BibTeX entry and citation info
15