Extracting timing data of both parent tier and time divided child tiers

Hi!

I am trying to find a way to extract timing data related to two tiers:

  • one parent tier ‘A’
  • one child tier (time subdivision as type) ‘B’ (edited CV: yes/no)

The question I’d like to answer is:
On average for each occurrence of A, to what extent do yes values for B overlap with annotations of A?

It would be easy for me to do the following:
sum of durations of B yes values / sum of A duration * 100

But this would be little fine-grained and not informative as to whether sometimes B does not take the yes value at all or for a very short/long time.

Would someone have an idea as to how one could compute this despite A and yes values not being time aligned?

Thank you!

Hi,

If I understand correctly what you wish to achieve, you could have a look at the options of the Annotations from Overlaps function (with filtering). See this section of the manual. Let me know if this does not solve the problem.

-Han

Hi,

Many thanks for your answer, Han.

I am somehow unable to click on the link but I have gone and read the instructions related to the creation of Annotations from Overlaps. I am not sure how to proceed: I have tried creating a tier where the value of the resulting overlapping annotations is specified as overlap duration. My issue is that I’d like to subtract the duration of no’s in B from the duration of each corresponding annotation on the parent tier A. The point will be to calculate the mean duration overlap between A and B’s ‘no’ values.

Thank you again for your help!
Sébastien

Hi Sébastien,

I made a typo in the link, that’s why it didn’t work, sorry about that. I corrected it, but you already found how to proceed.
If I’m correct and you’ve got a new tier with only the B ‘no’ overlaps with ‘A’ annotation, then you could go to View->Annotation Statistics and extract the necessary information from the Annotations II tab and/or the Tier tab. The total durations would allow to calculate the average or mean of the actual overlaps.

You mention subtraction of the ‘no’ overlaps and that could be achieved by the Tier>Annotations from Subtraction…, allowing to subtract the overlaps annotations from the ‘A’ annotations.
This can all help to get averages. ELAN has no single function to calculate the median of overlap durations. You could try the File->Export Multiple Files As->Annotation Overlaps Information…, select tier A and the overlaps tier and calculate overlap duration ratios per overlapping annotations etc. in a spreadsheet application?

Hi Han,
That was very helpful. Thank you!