Who knows the answer?

  1. How can you increase the rate of automatic classifications (less unknowns) for unstructured documents without sacrificing their reliability level?

Select one:
a/ By lowering the confidence level of the classifier
b/ By training the classifier manually in the validation step during the actual operation of the robot.

  1. How can you improve the reliability of the automation (less errors)without sacrificing the operational costs (efficiency)?
    Select two:
    a/ Use manual validation for all automatic classifications
    b/ Raise the automatic confidence level of classifications
    c/ Build verification automation for checking the extractions from other systems (ERP, CRM etc)
    d/ Fix the classifier/extractor rules manually based on erroneous classifications

I would also appreciate if, in addition to the answers, you can provide links to resources where I can study this particular topic in more detail.


Welcome to the community

  1. b is correct I believe …Becasue if we lower confidence the classification rate might increase but the classification might not be correct…for any model more trainign gives more accurate results
  2. d - The whole purpose of validation centre and action centre is to intervene when there is low confidence and retrain on the new variations or erroneous variations

Hope this helps



1 - b. By training the classifier manually in the validation step during the actual operation of the robot, allowing the system to learn from real-world data and improve its classification accuracy without compromising reliability.

2 - b,c - As these measures can improve reliability by increasing the accuracy of automatic classifications and introducing automated checks for data extracted from other systems, while still maintaining operational efficiency.

To review more you can surf through research papers on ArXiv, Association for Computational Linguistics (ACL).