Aligning closed captions for CNN newsroom

[Przemyslaw Grabowicz]

Hi Alberto, Hi All,

First of all, thank you for developing and maintaining this software and its documentation. It looks great!

I'd like to have a better sense of what aeneas is able to handle and what not. Specifically, I'm trying to align a 10-minutes-long CNN newsroom audio with its closed captions, but I haven't been able to obtain good results yet. The transcript contains original CNN closed captions, so it's a high quality transcript with maybe just 1% of words misspelled. However, there are a couple of 15-second-long time windows in the audio that haven't been transcribed, which perhaps is the reason why aeneas is failing to align the text. Is this a big problem for aeneas? Is there a way to handle this kind of situations?

Currently I'm using this command to get the alignment map:

python3 -m aeneas.tools.execute_task cnn_sample.aac cnn_sample.txt "task_language=eng|is_text_type=plain|os_task_file_format=srt|is_audio_file_head_length=13" cnn_sample.srt

I'm attaching the input files here (10 MB). I'm looking forward to your suggestions about what I could improve to get better alignments.

More generally, does anyone here has experience with aligning this kind of TV news broadcasts with their closed captions? Do you think alignment for such broadcasts is feasible with off-the-shelf tools like aeneas? If so, then which tools are most suitable for this task as of now? (I've seen Alberto's summary of multiple forced alignment tools already, but I'm just starting my exploration. I'm especially interested in multi-lingual methods.)

Thanks,
Przemek.