search

elpaco-escience / scikit-talk

Scikit-talk is an open-source toolkit for processing collections of real-world conversational speech in Python. The toolkit aims to facilitate the exploration of large collections of transcriptions and annotations of conversational interaction.
Apache License 2.0
2 stars 0 forks source link

add eaf parser #36

Closed bvreede closed 10 months ago

carschno commented 10 months ago

I suppose this is the specification of EAF.

carschno commented 10 months ago

EAF parser: https://speach.readthedocs.io/en/latest/index.html

mdingemanse commented 10 months ago

The main trouble with eaf is that you don't know beforehand which tiers are going to contain the utterance content that we actually want to parse.

This means you'll need to have some kind of UI where either

  1. you let the user give a list of desired tiers (this presupposes the user has that list beforehand)
  2. you present the user with some sample content, allowing them to choose the desired tiers

After that, you would probably also need some checks that will enable the user to see whether they made the right choice, and redo as needed.

mdingemanse commented 10 months ago

todo for @liesenf: pick 2 openly available eaf corpora that we have parsed before, enumerate the desired tiers & show the parsing code

bvreede commented 10 months ago

One would be IFADV

bvreede commented 10 months ago

The main trouble with eaf is that you don't know beforehand which tiers are going to contain the utterance content that we actually want to parse.

This means you'll need to have some kind of UI where either

  1. you let the user give a list of desired tiers (this presupposes the user has that list beforehand)
  2. you present the user with some sample content, allowing them to choose the desired tiers

After that, you would probably also need some checks that will enable the user to see whether they made the right choice, and redo as needed.

@mdingemanse check the documentation! Any feedback/corrections, feel free to re-open this issue.

mdingemanse commented 9 months ago

docs are looking good!