Convert "word numbers" to their decimal representations

[motiwari]

This transformation converts "word numbers" to their decimal representations in sentences, e.g.:

There are three hundred twelve million, five hundred thirty four thousand, six hundred seventy two people in the United States and one in every two is female. -> There are 312,534,672 people in the United States and 1 in every 2 is female.

This is a rather nontrivial transformation (see code). It is something of a reverse transformation to PR#71 and PR#39

[motiwari]

Hi, this PR is still a work-in-progress; please do not review (UPDATE: ready now)

[rteehas]

Hi, this PR is still a work-in-progress; please do not review

Is this still in a do not review stage? The deadline for me to submit reviews is the 18th I believe

[motiwari]

Thanks @rteehas , still a WIP, will update shortly!

[rteehas]

Tagging @kaustubhdhole to see if I can hold off on reviewing this until it is in its final state

[motiwari]

Tagging @kaustubhdhole to see if I can hold off on reviewing this until it is in its final state

Thanks -- almost done!

[motiwari]

Hey @rteehas done! Apologies for the delay.

Please see the updated code, PR title, and PR comments. Let me know if you have any questions!

Perhaps surprisingly, this transformation was rather nontrivial. I hope it will add value as an augmentation!

[ashish3586]

Task TEXT_TAGGING is not applicable here because of a change in the number of words (Ex: I have two hundred fifty books --> I have 250 books.)

[ashish3586]

You can add tasks PARAPHRASE_DETECTION, TEXTUAL_ENTAILMENT

[ashish3586]

Please add a docstring to the class WordToNumbers.

[ashish3586]

Please add a robustness evaluation section here like PR #218.

[kaustubhdhole]

Also, you might want to add this to the README and mention a line or two how your transformation is different. https://github.com/GEM-benchmark/NL-Augmenter/blob/main/transformations/number-to-word/transformation.py

[kaustubhdhole]

@motiwari would you like to address the above comments?

[motiwari]

Thanks for the ping -- I'll address the comments this week

[motiwari]

Hi @kaustubhdhole and @ashish3586 , thanks for the comments above -- I've addressed all your comments except adding the robustness evaluation.

I just rebased this branch on main and am now getting the following error that don't look related to my code; do you know how to fix? It's for the robustness evaluation:

(base) ➜  NL-Augmenter git:(dateformat) ✗ python evaluate.py -t WordsToNumbers -task "TEXT_CLASSIFICATION" -m "textattack/roberta-base-imdb" -d "imdb" -p 20
Traceback (most recent call last):
  File "evaluate.py", line 3, in <module>
    from evaluation.evaluation_engine import evaluate
  File "/Users/motiwari/Desktop/NL-Augmenter/evaluation/evaluation_engine.py", line 1, in <module>
    from evaluation import (
  File "/Users/motiwari/Desktop/NL-Augmenter/evaluation/evaluate_ner_tagging.py", line 2, in <module>
    from datasets import load_dataset
ModuleNotFoundError: No module named 'datasets'

[motiwari]

@kaustubhdhole @ashish3586 @rteehas @sebastianGehrmann can you take a look at this?