You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The errors caught by Keith are produced by wrong or incomplete identifications. One is solved with #44, but the other one cannot be solved because we expand only to "O" characters:
I'm guessing this error is because we never saw Month+Year in the training data, always Month+Day+Year. If that's the case, the right fix for this is probably to generate more artificial training data to provide such examples.
I could imagine a bunch of heuristics to fix this (e.g., never let a year be only 3 digits), but they all seem hacky, and don't get to the root of the problem.
The errors caught by Keith are produced by wrong or incomplete identifications. One is solved with #44, but the other one cannot be solved because we expand only to "O" characters:
In this case, "2" is identified as
Day-Of-Month
and "017" asYear
. This is why the resulting intervals is[0017-10-02T00:00, 0017-10-03T00:00)
.The text was updated successfully, but these errors were encountered: