|
--- |
|
tags: |
|
- politics |
|
- summarization |
|
- climate change |
|
- political party |
|
license: afl-3.0 |
|
language: |
|
- en |
|
- es |
|
- da |
|
- de |
|
- it |
|
- fr |
|
metrics: |
|
- rouge |
|
--- |
|
|
|
## Facebook/bart-large-cnn model fine-tuned on 7k political party press releases from 66 parties in 12 different countries. The model is trained to summarize each press release by providing the primary issue of the press release, the position of the party on the primary issue, and a 1-2 sentence summary. |
|
Training Data primarily consists of GPT-4 responses asking for summaries of the press releases. Small modifications were also made to the summaries from GPT-4 when validating the responses. I also made all the training text summaries lower case by accident (oops!), so outputs are lowercase. |
|
|
|
**Note** The model is pretty good at identifying the primary issue of any text, but it'll refer to the author of the text as 'the party' and summarize the "position" of *the party* as such. |
|
|
|
**Countries included in Training Data** = ['Italy', 'Sweden', 'Switzerland', 'Netherlands', 'Germany', 'Denmark', 'Spain', 'UK', 'Austria', 'Poland', 'Ireland', 'France'] |
|
|
