Closed AmyOlex closed 6 years ago
This was due to the format of hh:mm that is frequent in the clinical texts. Our methods were looking specifically for the hh:mm:ss format and were not catching the more common hh:mm format. I edited the code to also identify the hh:mm format, or the h:mm format, and now we are getting over .90 F1 for this entity. I think this is also messing up the HourOfDay as it is not getting all the subintervals it is supposed to.
We are missing most of the Minute of Hour entities in the cancer corpus.