Columbia Games Corpus - Question Identification Guidelines
-
Utterances that request an answer should be marked as 'q'.
- Utterances should be labeled as 'nq' if
a) there is something in the utterance that is plausibly questionable from the context, and
b) the utterance allows, rather than requests, an answer.
- Normally q's and nq's should begin and end at a pause, unless there is some
overriding reason to mark them otherwise. (There can be pauses within q's or nq's.)
- Label any strange cases with 'q?', 'nq?' or '?'.
How to label questions using Praat
- Each TextGrid file has four interval tiers: words and questions
for each speaker.
- Add one interval (labeled 'q', 'nq', 'q?', 'nq?' or '?') for each question in the corresponding tier.
The interval should start at the beginning of its first word, and end at the end of
its last word. Leave empty the intervals in between questions. Example:
- Before starting, play around with the settings in File » Preferences
in Praat's Edit window. You may find that helpful. For example, "Paint intervals
green whose label... IS EQUAL TO ...the text: #" paints all silences green in the
words tier, as shown in the example above.
- The best way to create intervals is to click on the boundary of the word (left
boundary of the question-initial word, or the right boundary of the question-final
word), which shows a little circle in the question tier. Clicking on this circle
creates a new boundary that coincides with the word boundary.
- Please do not modify the words tiers.