microsoft / okapi Goto Github PK

Hi,

this is very interesting work! Thank you for effort for collecting the data and making it easily available!

While the paper doesn't say this, the API calls are, to my understanding, essentially trees and the semantics should be the same across the different domains. Then, according to my understanding, the trees are linearized in the tsv files with a mixture of prefix, infix and suffix(?) notation. But then some things get me confused: For example, why does COUNT appear at the end of the (linearized) representations in the document domain but occurs more in the front in the email domain? How do we know what exactly is counted? Is my assumption wrong that the API calls are trees and maybe rather a set of tuples?

The other thing I noticed was that there are considerable number of duplicates in the data, e.g. https://github.com/microsoft/Okapi/blob/main/Dataset/Length_Split/document_domain/nl_api.dev.tsv#L761. Is that on purpose?