aikuma / aikuma-ng

Speech annotation web app for regular folk
22 stars 13 forks source link

How to mark sections as background noise or 'to be translated later' #117

Open nyborrobyn opened 8 years ago

nyborrobyn commented 8 years ago

Is there a way to indicate that a section is just background noise or chatter? Related to this, I don’t speak enough Mien yet so the background noise may sometimes be intelligible speech that I don’t know the meaning of. Is there a notation that should be used for “come back and annotate later”? I am using “(background noise)” or “(chatter)” but don’t know if that’s good practice. Same question applies for translation/transcription.

Lingomat commented 8 years ago

Ahh! I had this notion to provide the tag selector on the transcription interface so you can actually add tags to individual segments. These could be indicating speakers, language, or anything you like #unintelligable. It fell by the wayside in this spin of the annotation UI.

I'm rammed for the next week and a half though so it might take me a while.

First, perhaps you can help with the design of it? The starting point is the tag selector on the session view. So in essence you can type anything you want and it will offer to autocomplete any previous tags, so you would only need to type a couple of characters usually. You can play with the current tag selector on the session view.

I'm wary of breaking a nice work flow though, so I had in mind some streaminglining. If you use a tag on an actual annotation region instead of a session, then it's reasonably that you might want to use it again. So in this case we could show a numbered list of previously used region tags. Imagine you have (0) #Robert, (1) #Mary, (2) #Unintelligable, (3) #NeedsReview and so on.

These tags would actually be displayed and some sort of key shortcut would quickly add them without moving your focus like, say, control-2, and the unintelligable tag is added.

Thinking about it. It's probably better to use the person selector for people, and perhaps use the same kind of system. If we know we have two particular speakers, display them and offer a way to click and short-cut keys to add them to this span.

What do you think?

118 is necessary to implement.

nyborrobyn commented 8 years ago

Ok, would love to help with the design. By 'tag selector on the session view', I think you mean sort of the dashboard, or the whole view of a particular recorded segment including image, sections showing annotations, location, etc.

For tagging segments, I think you'd need to be inside a particular annotation, right? Rather than tagging the main session?

My impulse is to go into the annotation, right-click on a segment that's annotated, and be able to select "add a tag" or just have a box with open field that I can type in. In the event there are existing tags for this set, like you said, I want to choose from those that I've already added.

I like the idea of not messing with your focus using a keystroke (you suggested ctrl-2).

Yeah, seems reasonable to rely on person selector for people so you can associate different users with annotations/sessions.

What's the most helpful way I can give suggestions about this?

Lingomat commented 8 years ago

Great, no surprises in how you'd expect it to work. I think we just need to get it done and get your feedback on it.

I think we need a way to launch an experimental UI actually. Perhaps we can put a switch on settings. This likely won't appear for a couple of weeks at least.

nyborrobyn commented 8 years ago

Okay, I'll stand by until I can be of help.