PROJECT: Update outdated community workflows

[lkiesow]

The current community workflows are partly outdated and lack modern features by default (e.g. subtitles) while still including old, mostly unused features (e.g. OCR). They are also provided in different formats (some YAML, some XML) and partly still contain artifacts of the old admin interface.

We want to convert all workflows to one format and rework them to provide a better user experience to adopters by default.

Osnabrück University may fund this (work or money), but that's not yet ensured.

Additional information

• Responsible Person: Lars Kiesow and Rüdiger Rolf from Osnabrück University
• Additional Stakeholders: –
• Related Feature Requests and Issues: –

[oas777]

@lkiesow, you considering OCR one of the "old, mostly unused features" makes me wonder whether new custom workflows wouldn't include this. And what would this mean for publications towards Tobira, e. g.?

[mpmuc84]

Just noting it here: I/we would be interested in joining this effort, at least by testing, but also discussing the workflow structure and/or helping to rework the provided workflows.

[lkiesow]

@oas777, I would remove it since we have far better timed metadata by now: subtitles. But removing it from the default workflows also doesn't mean you can no longer use it. You just need custom workflows.

[oas777]

That I do understand. It just boggles my mind why you would abandon such an important feature in what is (to become) our default workflow - anyone using Tobira would have to have custom workflows? That can't be good for adoption in the sense of creating hurdles.
Also, if you point to subtitles : My understanding is that OCR is "unused" because there are no search features for the LMS or OCR results are not (considered to be) required. If you build such features for subtitles or if a time-based search using subtitles becomes a feature request, it would be foolish to omit OCR. Would be a pity if you then require everyone to add custom workflows just to use a standard time-based search.

[mtneug]

Not sure what Tobira is offering specifically regarding ORC, but regardless, we never found the results convincing / of good enough quality to be useful. Slides are tricky to get right. Often there is repeated text around the main body, and yes, in theory you can crop around that. But optimizing for one slide template doesn't solve it for all other templates. I agree that subtitles are much better timed metadata. The quality is usually very high and therefore its usefulness.

[oas777]

It's not just about the technical quality of a metadata, it's also about its usefulness: Students don't just remember words spoken, they also remember slides and their content. Therefore, it's key to combine results to serve different users. Relegating OCR to custom workflows would somewhat ignore that.

Not sure what Tobira is offering specifically regarding ORC, but regardless ...

Maybe time for a quick update on our custom distribution channel before you jump to conclusions? Cf. https://explore.opencast.org/~search?q=matthias.

[mtneug]

Maybe time for a quick update on our custom distribution channel before you jump to conclusions? Cf. https://explore.opencast.org/~search?q=matthias.

What do you mean with "custom distribution channel"?

I tested a bit with the Tobira search and see your point now. Yes, adding OCR results to search seems reasonable to me. My preconceived opinion stems from the usage in the Theodul Player, where the OCR was presented as is (and thus often incomprehensible) below the player. Because of this, I couldn't see the point of OCR. So now in Tobira you are basically treating it as a "word cloud" (not quite, but similar) that is enriching the search? You therefore don't care about repeated text from every slide, correct?

[oas777]

I'm glad you see the advantages. I cannot speak technically, but repetitive text was not an issue when we discussed and tested search for Tobira.

What do you mean with "custom distribution channel"?

I consider Tobira to be / to become just that. "But what about the LMS?" I hear you wonder. Well, they are not part of Opencast. I know they are important with respect to our role in teaching and learning, but from a (marketing/product/USP) perspective Tobira is (going to be) just that.

BTW, if we digress any further, that Matthias guy from Münster is going to reprimand us.

[lkiesow]

Quick update: First step is done. Our workflows now all have the same format: YAML. This didn't change any functionality.