Wip confidence levels

[JamesSG2]

• Added a 'low_confidence_rejection' function to the backend.py (with comments) which combined with other changes I made, finds the confidence level for each prediction, and removes the words and their respective letters that are below the minimum confidence Level (currently 55% confidence -- is adjustable).
• Can discuss further in the next meeting why a CL works for words but can't work on the character by character basis.
• Currently this system rejects about 50% of all incorrect words, in exchange for rejecting ~5% of correct words.
• I also spaced out the "Analyze" results to make them easier to read.
• A system for alerting the user if all instances of a letter were rejected (or never existed) should be the next step.
• Also no need to delete Wip-confidence-levels branch after merging; I can rename it and use for my next project / feature

[rover-t]

Looks good. Definitely an improvement over what it was before. The only issue I ran into is that I had to pip install pandas to get line 35 output_type='data.frame' to work. I would add that to the requirements.txt file so that it is listed on the requirements page.

I also noticed that the low_confidence_rejection method directly acts on pytesseract_list rather than returning the corrected list separately. This works, but it might be better to return an updated list rather than acting on an existing list. I'm not really sure though, so I'll leave it up to you if you think we should change it.

Also worth playing around with - is there a way to run the regex correction I wrote before running your function? The more original text there is the more effective the regex will be. If that's not possible it seems great as is, but its just something I thought of.

[JamesSG2]

I'll go ahead and merge wip_confidence_levels to dev now since it's been reviewed and I just made the suggested changes.