Our bot needs to recognize numbers from sentences. We’ve incorporated Facebook Duckling for this task. But in the result we’ve got many false positives from sentences like:
“xxx one” → “lucky one”, “better one”, “second one” and so on.
We’ve applied regex matching to discard a bunch of them, but it is not very robust solution… Could you provide a hint where we should find something more general and not “case specific”?
Beside that Duckling is brilliant and we don’t want to discard it. Possibility to handle sentences like “this Christmas” is amazing.