Tutorial for term 2 week 9

TUTORIAL 8: for Semester 2 Week 9

Annotating Dialogue

This week you will discuss how to annotate video recorded dialogue and you will practice on this video → Link to tutorial7-dialogue-video.mov

First of all come up with a plan how you would annotate speech in general:

1. What is the purpose of my annotation? e.g.

Do I need annotations of my data for therapeutic reasons?
Do I need annotations of my data for subtitles in a film?
Do I need annotations of my data to teach a machine, computer, speech recognition?

2. What are the differences between the above mentioned purposes? e.g.

Orthographic vs. phonetic annotations.
Broad vs. narrow annotations.
When do we need to annotate background noise?
Distinguishing speech from other human noise.
Can we annotate all information in our speech data?

Next, go back to your video and think how you would annotate this specific data.

3. What do we need to consider when we annotate tutorial7-dialogue-video.mov? e.g.

How many labels will you need?
What will your set of labels look like?
How do you annotate speech when both participants talk at the same time?
How do you label laughter?
Do we need to annotate silence in dialogue?

4. Video data contains more communication channels than speech data only. How can we annotate speech, gesture, facial expression?

Will we have to come up with a different annotation plan for each different communication channel we can find in video data, e.g. for speech, gesture, etc.?

During the tutorial you will discuss annotation schemes you could use. You will think of labels you can create to mark all communication events in your video data.

For help and annotation you can find an annotation scheme here (link to http://www.phonetik.uni-muenchen.de/Forschung/SmartKom/Konengl/engltrans/engltrans.html).
This is a transliteration manual/annotation scheme use in the project SmartKom for German language but with English explanation.

Home : Teaching : Courses : Hc1