Data Extraction and Sentimental Analysis from Equity Research Reports at Scale

Please read 'requirements.txt' for details.

Demo of CV module can be found in ./notebooks/. To process files in batch, please follow the steps:

• Firstly preprocess data, run code/cv/convert_pdf_to_jpg.py
• To test the public models, please:
  • Run code/cv/download_models.jpy
  • Run code/cv/parse_layout.jpy
  • Call functions in code/cv/evaluate.py. notebooks/evaluate.ipynb is recommended to see the usage.
• To finetune model, please see codes in code/cv/layout5_detectron2.ipynb.
• Finally, run code/cv/OCR.py to extract characters, but API_KEY and API_SECRET is required.