atlasharry
commented
4 months ago I have gone through the first five repos:
1) Entity Linking 2022 did not use any 2020 training data 2) Race of Focus uses
- fb_2020_140m_adid_var1.csv.gz that is not on our GitHub yet. I believe this is one of the main 140m final data outputs from fb_2020. 3) Aspect-Based Sentiment Analysis uses
- fb_2020_140m_adid_text_clean.csv.gz 4) Ad-Level Party Classifier uses
- fb_2020_140m_adid_text_clean.csv.gz
- fb_2020_140m_adid_var1.csv.gz 5) Party Classifier with Unique ID uses
- fb_2020_140m_adid_text_clean.csv.gz
- fb_2020_140m_adid_var1.csv.gz
In conclusion, those 5 repos only use 2 of the main output data from fb_2020: fb_2020_140m_adid_var1.csv.gz and fb_2020_140m_adid_text_clean.csv.gz. The link to each file above would also redirect to the files and lines where the file was mentioned in the code.
a-jacewicz
furkmak
SebastianZimmeck
As part of our classification repos, we would need to provide any training data we use, including data from 2020 cycle. Harry, Aleks and I will go through each of the following repos to note if we use any 2020 data for training that we need to upload to Figshare due to its size (larger than 100 mb):
To reiterate, we do not need to worry about any 2020 training data that are already uploaded to GitHub. We just need to check if a training script uses a 2020 training data that is not readily available on GitHub. If you find any files that we need to know, just make a comment here once you go through all repos above. Feel free to share the workload as you see fit.