Closed dcanones closed 3 years ago
Hello, do you have more concrete examples and details (e.g., which CHILDES dataset you used, which IPSyn rules)?
I've just looked at the IPSyn code. All the 56 items are based on the dependency graph of a given utterance and rely on morphological and syntactic/semantic information (e.g., part-of-speech tags), and there doesn't seem to be anything related to punctuation marks.
I'm closing this ticket for now -- we can reopen it if further info like which IPSyn items specifically are problematic and/or relevant CHILDES data is available for debugging.
Hi!
It seems like the IPSyn level calculation is not accurate. It takes punctuation marks (e.g. [.]) as a full valid word, triggering some of the IPSyn rules and increasing its value artificially.