Open kajad opened 2 years ago
I find this feature to be very interesting. The problem is similar to ours.
I'm not sure if sentence-level resolution is the best option. Multi-sentence resolution might be better (in our experience, taggers prefer 30-60 seconds of audio).
Maybe this feature will address your use-case: https://github.com/inception-project/inception/issues/5011
It allows importing a ZIP file containing a document in HTML format and optional media files, e.g. audio files. In that way, you can have a document that e.g. contains a transcript of the audio as well as the audio itself. The annotators could then annotate the transcript while listening to the audio.
We have several projects that deal with spoken data annotation, i.e. annotations performed on written transcriptions of speech recordings. In our experience, the annotators perform much better if they have the recordings at hand when doing the annotations on transcriptions.
Would it be possible to add this feature to INCEpTION, perhaps by using the image URL linking feature to link to URLs of audio recordings instead?
We are primarily interested in linking to the sentence-level recordings (i.e. one URL per sentence), but I imagine other users might be interested in linking to paragraph- or document-level recordings as well.