Variable deletion rules for (Scottish) l-vocalisation?

[Sophie Holmes-Elliott]

Hi,

I am working on a project investigating bidialectalism in Buckie, Scotland. We have run most of the corpus through FAVE-align and it's done an excellent job, seems to have coped really well with the data even though the accent (and grammar) is quite divergent from the varieties it was trained on.

One thing I noticed when inspecting the alignment on the text-grids was the variable alignment of ING/IN' - a really useful feature for our data as it's likely to be a feature we look at. I have two questions related to this:

1. Is there a script I can use to extract this variable along with other info? For instance: whether velar/alveolar realisation occurred, the word, and the surrounding phonetic context?

2. Is it possible to do this for other features? For instance, in this variety of Scots, l-vocalisation operates as a categorical phenomenon where the /l/ is realised or not (i.e. ball/baa call/caa, all/aa etc etc). So it would be really useful to be able to use a similar variable deletion option and I just wondered how this could be written into the program?

Obviously, as this feature is lexically conditioned, we would need to provide an exhaustive list of which words were variable, but I just wanted to check that this was feasible before we compiled one.

Thanks!

Sophie H-E