I am unable to lock segments

[Trsa993]

Hello,

I'm trying to align text with timestamps, but I'm unable to lock words that starts with lowercase letters.
So I don't want to split words to new segment that starts with lowercase letter like here:

{
{
"text": " Somewhere in between the earth and the stars I know there's gotta be a new place to start Where I can catch my dreams and not lose your",
"start": 120.62,
"end": 139.02,
"words": "..."
},
{
"text": " heart Out there somewhere in between the earth and the stars",
"start": 139.62,
"end": 150.06,
 "words": "..."
}
}

And my code is this:

lowercase_letters = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z']

result = stable_whisper.alignment.align(model, 
                                                                str(path),
                                                                text,
                                                                regroup=False,
                                                                language="en"
                                                                )
result.lock(startswith=lowercase_letters, case_sensitive=True)
result.regroup()
result = result.to_dict()

I'm probably doing something wrong, but I can't figure out what.

Thank you.

[Trsa993]

Ok, so I saw that instead of startswith, I need to put endswith. Now I have other problems, but I hope I will figure out.

[jianfch]

They failed to lock because words start with a space. It should work for e98c3d6.
Note that for lock(), only right is True by default. You might also want left=True for your case.

For example, locking earth in the earth and
Locking the right only prevents splitting/merging between earth and and
Locking the left only prevents splitting/merging between the and earth

[Trsa993]

Thank you for the answer!

Also is it possible to unlock words, for example something like this:

result.regroup(regroup_algo='cm_sp=.* /。/?/？/,* /，_sg=.5_mg=.3+3_sp=.* /。/?/？')
result.lock(endswith=uppercase_letters, case_sensitive=True, left=True, right=False)
result.merge_by_gap(2)
result.lock(endswith=uppercase_letters, case_sensitive=True, left=False, right=Fales)
result.lock(endswith=lowercase_letters, case_sensitive=True, left=True, right=False)
result.split_by_gap(.3)

[jianfch]

There's no method of unlocking specific words, but you can unlock everything then relock the specific ones you need. You can unlock all with result.unlock_all_segments().

[Trsa993]

One more question. In the code below, when I use split_by_length, is it possible to use lock, because it will not lock lowercase words. It would be good if it splits the first word that is not locked after max_words.

result.clamp_max()
result.lock(startswith=lowercase_letters, case_sensitive=True, left=True, right=False)
result.split_by_gap(.1).split_by_length(max_words=30)

Thanks in advance!