Hello
I have used ELAN to code parent and child speech between playing with different toys. I want to calculate MLU, MLT and TTR in CLAN. I have constructed my tier hierarchy as play context (the toy type) as the parent tier. I selected Play Context as the parent of you ELAN hierarchy, and the parent is the one thing that one cannot change later. I would need to have two parents: ADU and CHI. The other tiers would be renamed to things like typ@CHI. I can’t figure out how to do this. Perhaps someone could help me?
If I change the name of the tiers would this be enough? Would the hierarchy still effect the files?
TIA
Hello,
There are two (almost identical) ways to change the parent of a tier in ELAN. The first one is Tier->Change Parent of Tier..., the other one Tier->Copy Tier.... In both cases the tier(s) is/are copied, the latter function has a few more options, so that might be the better choice.
I’m not sure if this change in hierarchy is required for the conversion to chat format (I assume you’re going to use CLAN’s elan2chat command for that?) and subsequent calculation of MLU etc., but I guess it is.
-Han
Hello Han
The hierarchy change isn’t necessary for the conversion, but it necessary for the computation of MLU since my parent tiers are not speaker tiers. In order to calculate the MLU, etc. I need to have parent tiers which are speakers.
This will be problematic since I am interested in the MLU between different toys (my parent tier) I am at a loss as to how to organise my data here! I need to think…
Hello Han
I am trying to change THE parent tier, not the parent of a tier. In order for MLU, etc. to be calculated the parent tier must be a speaker tier. My parent tiers is which toy was being played with…Is there a way to change this? I am so sorry for being a nuisance
Hello Leanne,
Yes, my reply was maybe not very clear. If I understand correctly, you now have the ADU and CHI tiers (maybe they have different names still) as dependent tiers of the Play Context tier. The speech tiers themselves may or may not have dependent tiers as well. What you can do is the following:
- open the
Tier->Copy Tier... dialog
- in the first step select e.g. the
ADU tier, select the Copy depending tiers as well checkbox
- click
Next
- in the second step select the
Transcription (No Parent) row in the tier tree (this means the copy of the tier will be a new top-level, parent tier)
- click
Next
- in the third step you’ll have to select a tier type. (This will be a different type from the type the original tier has, because the copy tier will be a top level tier.)
- click
Finish
There will now be a new top-level (parent) tier, named something like ADU-cp (depending on settings and actual original name). Repeat with the CHI tier etc. Now Play Context and ADU and CHI are all top-level (parent) tiers. Maybe you want to delete the original depending tiers.
I don’t know if CLAN has options to calculate MLU etc. per toy (by cross-correlating tiers). If not, you could, in ELAN, select each Play Context annotation one-by-one and choose File->Save selection as .eaf... to create separate ELAN file per toy. These could then be converted to CHAT one-by-one to calculate MLU per file. This would be an awful lot of work, obviously.
Needless to say that you have to make sure you have safe backup copies of the files you start out with!
Hello
Your advice many months ago was perfect and it has served me well, until now. I am performing the exact procedure you outline above (and I have done this 30 times with 30 different data sets) only to be met with an issues in CLAN. It now tells me illegal speaker characters are present on line 7 (for the 5 different data sets I have tried. I have emailed Prof MacWhinney (the creator of CLAN) to see if he can see the issue, but I also thought you may know the reason? I literally computed the TTR for one set and the next, wouldn’t allow it! I am still using the old version of ELAN too…
I am so sorry to trouble you and ever so grateful for your knowledge.
Leanne
Hello,
That’s too bad. I assume Prof MacWhinney can point you out which speaker characters are problematic, otherwise you’re welcome to send an example .eaf file (and possibly the CLAN exported/converted version of it, if conversion produced anything) to me (han.sloetjes AT mpi.nl) and I will have a look.
If it turns out to be a matter of changing tier or participant names, there is a multiple file (“batch”) function for that.
-Han
