This is a website to generate political canidate speeches using Open AI's GPt-2. It runs on top of gpt-2 cloud run. The model training was done from speeches found on cspan for each canidates via the text transcripts. The backend runs on google cloud run. If you would like to learn how the infastructure works please look at https://github.com/minimaxir/gpt-2-cloud-run.
Under mining the code to scrape cspan videos for the transcrips can be found: STEPS TO RUN PROGRAM:
- Install all packages
pip3 install -r requirements.txt - cd into the
miningfolder to run the transcript_scraper - run
python3 transcript_scraper.pywith the specified arguments
To see what arguments need to be passed in, type in python transcript_scraper.py --help
To get a url for the c_span_search_url input:
- Go to C-SPAN's main website and search for a term, then click on the videos button just under the search bar to only get video results.
- Include any filters you want in the search
- (Optional) Click on show 100 next to
showing 1-Non the right of videos sorted by - Copy and paste your full url (with the www) in the browser into the program
If all videos are to be concated into one file, the file name will show up as all_transcripts_concatenated.txt
Otherwise each file name will be speech name
example python3 transcript_scraper.py -c_span_search_url https://www.c-span.org/search/\?empty-date\=1\&sdate\=\&edate\=\&congressSelect\=\&yearSelect\=\&searchtype\=Videos\&sort\=Best+Match\&text\=0\&personid%5B%5D\=34\&formatid%5B%5D\=55\&show100\= -is_save_to_one_file True -add_video_name_to_file False
NOTE: If you want to download just one video's transcript you can you the function transcipt_scraper in here and just give it the video's url