Computing global gamma for a corpus of sentences/instances

[vestedinterests]

First of all, a great package, I am really happy that gamma exists as a measurement at all and also about a well-documented python implementation!

I had a brief question; say you are using this for an NER task. Your whole corpus might then lots of individual sentences which are annotated. I am now wondering how I'd best compute a global gamma for the whole corpus.

1. Reading the documentation, it seems that using the CLI, I could have each sentence in a file, then batch analyse them, have individual gamma measures per file and then report SD of gamma, lowest and highest values.
2. Or I could add each sentence after one another, meaning that token 3 in sentence 3 perhaps has the token-position 12, since I would treat is as one giant annotation task, and then have a gamma computed for the whole corpus.

I seem to see both approaches used in papers citing your work, most without code-sharing however; I was curious if you have a recommendation which approach makes more sense. Thanks a lot!

[hadware]

Sorry for the extremely late answer. First of all, are you still working on that topic? Have you found an answer to your question?

I can look into it if you're still interested.