Part of parent segments are lost when they are divided in their children about similarity-annotator HOT 6 OPEN

@Rafael-Caro Do you mean the right-most part? From the third tier of your figure, I can see the left-most boundary is ok. Maybe it's because you removed the silence segment? The silence segment happens to have the right-most boundary of its parent segment?

from similarity-annotator.

@Rafael-Caro @oriolromani Ok, I confirm this issue. This only happens in segmenting reference sound.

from similarity-annotator.

Sorry @ronggong , you're right, is the rightmost boundary, my fault!

So, in order to clarify the issue, I have started from the beginning, have deleted all the tiers and started again, and I'll paste a pic of every step:

I started annotating the reference sound.

1. Opened tier 3 (parent) and created this segment:
   

2. Opened tier 4 (child) and the segment is inherited. I click on it to put the cursor in the place where I want to divide it and instantly appears a new boundary (#97). I deleted it, then the cursor stays as usual, and click "i" to create a new segment. Eventually I get these three segments. The space between the first and second one is done just by dragging the left boundary of the second segment, because although that boundary is created by clicking "i", it is not tied with the right border of the first segment.
   

3. Opened tier 5 (grand-child) and the segments from 3 are inherited. I enable the spectrogram and zoom 4x (this works perfect (except for #99) and is really, really useful, kudos!), and start dividing those segments. In the process, I modify some of the boundaries inherited according to the spectrogram by dragging them. #97 and #98 still happens. Here'es the result.
   

4. Ok, until here everything perfect. But when now I open tier 3, that's what I find (compare it with the pic in step 1). The modifications I did in the inherited boundaries by dragging them in tier 5.1 have been propagated correctly to this tier:
   
   When I open tier 4:
   
   Tier 5.1 keeps the segments I did in step 3, and tier 5.2, synchronized with 5.1, have the same segments:
   
   However, when I open 5.3 (synchronized with 5.2) and 5.4 (synchronized with 5.3), this is what I find:
   

Then, I repeated the steps 1 to 3 in the target sound, with the same reults as shown in step 4. #97 and #98 also happens. The only difference is that when I insert a boundary using the key "i", the new two segments are tied by this boundary, that is, when I drag it, the two segments are modified, and no space appears like with the reference sound. So the space between the first and second segments in tier 4 in the target sound is done by creating a new segment that covrers the silence and then deleting it.

Then, I continued by segmenting tiers 6 and 7 in the reference sound, and I guess you can imagine what happened just by looking at the images:
Tier 5.1 (compare it with the pic about tier 5.2 above, it should have those segments):

Tier 6:

Tier 7:

Just to be clear, tier 5.1 and 6 (and also 3, 4, 5.2, 5.3 and 5.4) should cover the same "yellow space" as tier 7, which was the last annotated one.

Finally, I tried out tier 8, which is a timestamps tier, both in reference and target sounds. It works very smoothly, but only in case of the reference sound, the timestamps cannot be dragged before storing the annotations, and once stored, if I double click on a timestamp and drag it, a segment is created (#102). Again, only when annotating the reference sound, annotating the target sound works perfect.

That's all. Sorry for the very long comment. I really hope is helpful. And again, I use Windows 10 and Chrome 56.

from similarity-annotator.

By now we can overcome this issue by first creating the segments, storing them and then modifying them if needed afterwards

from similarity-annotator.

Great news! I did it in that way and it worked! So you don't have to worry much about this, thanks a lot! I only tried the reference sound though, I'll try the target sound tomorrow.

from similarity-annotator.

Ok, in any case @andrebola and I are working on it

from similarity-annotator.