-
Notifications
You must be signed in to change notification settings - Fork 4
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Integrate TAC NIL clustering measures #39
Comments
First cut at 3497f5e I will circulate after tidying and adding tests. |
Reviewing your changes to the output of coref metrics. You assume that all metrics can be calculated as a function of tp, fp, fn. I don't think this is true of the Cai and Strube variants which account for differences in extracted mentions (i.e. "twinless mentions"); I'm not sure if it's true of entity ceaf (haven't cogitated sufficiently); and it's certainly untrue of BLANC and its family, since BLANC calculates the macro average of the coreferent and non-coreferent precisions, recalls and F1s. We could treat BLANC, at least, differently, but for the others it might be better to allow a different numerator for each of precision and recall. |
I have updated the matrix class in evaluate as discussed. Coreference metrics now return p_num, p_den, r_num, r_den. Call _prf on this output to get precision, recall and f-score. |
Won't work for BLANC, but is a reasonable generalisation of the others. On 25 May 2014 14:52, Ben Hachey [email protected] wrote:
|
No description provided.
The text was updated successfully, but these errors were encountered: