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

SRT files - quick questions #40

Closed Morrizzzzz closed 1 week ago

Morrizzzzz commented 10 months ago

Back in the day (before streaming) I used .srt files for subtitles of movies/ series etc. Is that format still used? And how useful might this be in this project?

Maurice

mdingemanse commented 10 months ago

SRT is still a popular format for movie subtitles. One similarity with our kind of minimum viable format is that it includes timing information. A difference is that it doesn't do speaker identification. Probably because of that, no language corpora that we are aware of come in this kind of format, and it is not expressive enough to serve as an intermediary format. However, it could be a good starting point to learn about time-aligned linguitsic annotations.

bvreede commented 10 months ago

Would it be a good idea to consider adding it to the import types? This may open up sktalk for use by a whole different group of users?

liesenf commented 10 months ago

Yes, good to keep tabs on but low priority because .srt is rarely used in scientific work on conversation/speech.

mdingemanse commented 10 months ago

Afterthought: also worth noting that SRT has very different needs. For instance, subtitles don't need subsecond precision, and typically need to be on screen for minimally 1sec (e.g. here); many utterances in conversational speech are much shorter than that (see Figure 7L in same LREC paper).