Detecting Political Bias in Speeches and News Articles in the US Using Deep Learning Methods
Notes on Convote scraped dataset:
The test_set, development_set, and training_set were scraped from the Convote data_stage_three folder.
The data is stored in CSVs in the format text,party e.g. data,label.
In the raw folder is the the raw textual data.
The raw textual data has a lot of commas, so a semicolon ; is used as a delimeter between text and party.
The stripped folder has the same text stripped of commas, and a comma , delimeter is used between text and party.
Preprocessing is implemented in preprocess.py. In order to run the script, you need to install the huggingface library:
pip install transformers
I had some install problems that were solved by installing Rust, which it should be easy to find instructions for online.