Unclear Error message

[SamPassmore]

Hi,

I am using emuR to run a forced alignment pipeline. I understand emuR runs on MAUS - so perhaps this question is better pointed elsewhere, but I will start here.

I have been following the tutorial here:
https://www.phonetik.uni-muenchen.de/~jmh/lehre/sem/ws2324/Emuintro/03_forced_alignment.html#from-a-text-collection

I run this step (tweaked for my purposes):

runBASwebservice_all(SL_db,
                     transcriptionAttributeDefinitionName = "segments",
                     language = "eng-GB", 
                     runMINNI = FALSE,
                     verbose = TRUE,
                     patience = 3,
                     resume = TRUE)

Which returns the error:

bas_run_maus_dbi(handle = handle, canoAttributeDefinitionName = canoAttributeDefinitionName,  : 
  Something has gone wrong in the turn query...

Which I can't really decode. Do you have any advice on what I could check to see what might be causing the error?

Some additional information:
I have an emuDB of 336 bundles, where each bundle is approximately 400kb on average (maybe a little less).
I am using an Apple M1 Pro chip, on Sequoia 15.5 - which might not be useful information because I think this is run elsewhere.

Thanks in advance,

Sam

[MJochim]

Hi Sam,

hard to decode indeed. First thing I would try – not to solve the problem but rather to find it – is remove all bundles that exceed one minute and try again. This only makes sense when many of your bundles are less than one minute; I am assuming this is the case based on the 400 kilobytes you mentioned.

If you don’t know exactly which bundles are that long, one way to find them is this:

library(emuR)

# ... load SL_db ...

media_files = list_files(SL_db, fileExtension = "*.wav")

for (media_file in media_files$absolute_file_path) {
  data_object = wrassp::read.AsspDataObj(media_file)
  duration = wrassp::dur.AsspDataObj(data_object)
  if (duration >= 60) {
    print(media_file)
  }
}

The reason I would try this is that your message points to a problem in the Chunker stage, which is not used for recordings shorter than a minute. If this turns out to solve the problem, we can find a good way to deal with longer recordings, too.

[SamPassmore]

Unfortunately, I don't have any files over 1 minute. They are all around 10 to 20 seconds long (hence the small size).

I thought maybe the issue was that there was lots of files, but that hasn't seemed to have fixed anything.

I've managed to get things working using MFA now, so there is no panic. But I am interested in getting the process to work through emuR, if you have advice on what to look at.

Sam

[NPoe]

This is not a chunker problem, instead we seem to be using the “segments” level as chunk annotation. @SamPassmore Is that what you intend to do here? Are there any empty strings or NA labels in that level?