USE CASES -- Annotation of time-based media
From Open Annotation Collaboration
Contents |
General Session Notes
Notes Specific to Subscription Streaming Video Content
Notes Specific to Annotation for Ethnographic Video and Audio
Discussion Notes
- Tim :- Will, are the things listed on your segment description, the contents of the annotation? There may be differences in what annotation means to different people.
- Jane suggested that object identification ought to be automated. This is difficult on the Web based on codecs not identifying objects.
- Jim: we are still wrestling with the idea of how long we might display an annoation when playing the video.
- Jane: gave example of crocodile videos where scenes are annotated based on preprocessing with sensor data. This preprocessing is very important to improving the quality of the annotations.
- Jane: interesting application - using annotation tools to annotate a corpus that could then be used as input to a machine learning process.
- schuyler & jonah: looking at compound objects with mediathread
- looking at bringing in resources from outside sources like Alexander Street Press
- final product is a more traditional composition space containing links to media fragments from various media repositories (e.g. youtube, alexander street press, etc.)
- annotation is just a link to a constraint, if people are generating these annotations outside the system then OAC’s use would make it easier to bring them into the system
- annotation is a complex type of document, what is the role of OAC in terms of facilitating the modeling the body of the annotation (the document)
- jim: could be structured but should be machine-readable (i.e. rdf)
- questions about what that rdf should look like
- jane: some need for best practice guidelines
- tim: by way of illustration...sometimes this is multiple targets and sometimes this is an aggregate target; get the domain people to help with individual domains going into as much detail as they need to
- schuyler: will authors need to construct their rdf documents (with nlp, authors don’t have to construct all of those graphs)?
- manual vs human annotation: authors were asked to provide rdf-style annotation data for the use of training editors but the editors had to go back and fix the data because the authors weren’t very good at abstracting
- multimedia annotation and analysis tool: ELAN by the MPI for Psycholinguistics (ELAN is an offline tool, and ANNEX is its online viewer)
