UPPERCASE text

[adamajm]

I've found that all UPPERCASE text returns the [{Name}] result for almost every word. For example:
PANKETS 621 LLC - Business Phone: 919-256-2873 (JESSICA)
621 CHAPPELL DR
She would like to have BINS SENT OUT FOR UNITS 101-105
STATED THEY ARE BILLED FOR RECYCLING BUT DON'T HAVE BINS
SHE WANTED TO KNOW IF SHE COULD GET A CALL BACK ABOUT CHARGES FOR RECYCLING

Returns:

{{NAME}} 621 {{NAME}} - Business Phone: {{PHONE}} ({{NAME}})
621 {{NAME}} {{NAME}}
She would like to have {{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} 101-105
{{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} DON'T {{NAME}} {{NAME}}
{{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} GET {{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}} {{NAME}}

[deanmalmgren]

Yeah, that's the trouble with only using part-of-speech tags from NLP to detect names. PoS tags are flawed in that way. I think that using machine learning for name detection (#16) will be quite a bit better, or possibly using spaCy instead of textblob/nltk (#18).

The good news is that this isn't a terrible default. Although it is redacting a bunch of information that you wouldn't necessarily want to redact (ALLCAPS results in a "false positive"), it is better to be more conservative and not release any names if you can avoid it. I just added the precision label to this issue to be sure to look into this more carefully.

[thomasbird]

We've released some new detectors based on the spacy and stanford NER models, which should fix this issue. You can checkout the docs: https://scrubadub.readthedocs.io/en/stable/names.html