financial numbers reading

[kpanuragh]

This language data file support financial numbers like (5,555) 12,555

[Shreeshrii]

Make your PR for https://github.com/tesseract-ocr/tessdata_contrib instead of this.

Give comparative data about how and why this is better than existing data.

[kpanuragh]

Reading financial numbers is not that much accurate on normal eng.traindata, When reading data using normal eng.traindata it read numbers like

(4)

as

D

. I just find the issue on my personal project and it's not only in (4) also other numbers like (16) 1 etc, Some time the issue is present on large number also, so I just try to train model from eng.traindata, It only includes the financial number format(0123456789(),.) .

[zdenop]

@anuraghkp1 : as @Shreeshrii point out: can you please make pull request to tessdata_contrib repository?
And provide some details for it (e.g. regarding training - something like khmLimon.md for khmLimon.traineddata

[Shreeshrii]

@stweil can the PR be applied to tessdata_contrib directly? It will be useful to other users too.

E.g. see request on Shreeshrii/tessdata_shreetest#9

[stweil]

can the PR be applied to tessdata_contrib directly

Yes, that is possible. @anuraghkp1, could you please describe how you generated the new traineddata? Ideally it should be possible to reproduce your training process.

[stweil]

@anuraghkp1, ping.

[stweil]

@anuraghkp1, do you have some examples where the new model is superior to existing models? Would a white list of expected characters with an existing model achieve the same results as your model?