It seems these two phones are missing from the unilex-rpx phone list at the path listed in step 8.1 of the unitsel exercise. Is it fine to just manually add these two phones back before doing alignment?

As in title. For instance, with a text file like this:

du n no I.
that’s gone.
and somehow allegedly didn’t see anything I I.
you genuinely saw nothing.
like when I.
it’s a picture.

You get this output from Festival, where Lines 1&2, Lines 3&4, and Lines 5&6 are merged into a single utterance:

utt00000001 <{(1 d uu )}{(1 e n )}{(1 n ou )}{(1 ai )}{(1 dh a t s )}{(1 g o n )}> _BB
utt00000002 <{(1 a n d )}{(1 s uh m )(3 h ow )}{(0 @ )(1 l e jh )(0 i d )(0 l iy )}{(1 d i d )(0 n! t )}{(1 s ii )}{(1 e )(0 n i )(0 th i ng )}{(1 ai )}{(1 ai )}{(1 y uu )}{(1 jh e )(0 n y uw )(0 i n )(0 l iy )}{(1 s oo )}{(1 n uh )(0 th i ng )}> _BB
utt00000003 <{(1 l ai k )}{(1 w e n )}{(1 ai )}{(1 i t s )}{(1 ei )}{(1 p i k )(0 ch @ )}> _BB

I trained my data on a Dice machine, and there was an error saying “Disk quota exceeded”. I guess I am not allowed to have much disk space. But I only trained 800 sentences (I will need 2000 sentences in total). Is it possible to have more space?

Thanks!

I’m running the remote computer and keep getting this error – ‘Disk quota exceeded’ which is really frustrating as I’m unable to save anything – I have tried deleting some old files I no longer need . Is there any way to change my permissions or do I need to contact an admin?

I am trying to initialize and train a set of models on several users using -S. For some reason, after running the script the trained models in hmm1 folder have not been updated. If I delete them, it does not create new ones. Only by using the standard script that trains on 1 user, trained model files are updated/created in hmm1. I am not sure why this is, here is my script:

PROTO=5state
set -x

done

I had the script right on the recognise_test_data file before but I accidentally deleted it , I know I need to change something in the loop here but I’m not quite sure what for multiple test users :
# to use a single user’s data, for example “simonk”
USER_FILE=$1

# to use your own data, this would automatically set USER to be your username
# USER=${USER:-whoami}

# and this is the path to the data
DATA=${DATA:-/Volumes/Network/courses/sp/data}

shopt -s extglob

for F in ls ${DATA}/mfcc/test/${USER_FILE}_test?(_)??.mfcc
do
FF=basename ${F} .mfcc

echo
echo Recognising ${FF}

HVite -T 1 -C resources/CONFIG \
-d models/hmm1 \
-l rec \
-w resources/grammar_as_network \
resources/dictionary \
resources/word_list \
-S scripts/${USER_FILE}

done

Am I right in thinking we have 39 MFCCs per 25ms frame (12 mfccs, 13 deltas, 13 delta-deltas ?). Are we expected to explain the reasons why we used MFCCs in our report

I’m also a little confused about what our language model and acoustic model is in the context of this sentence about viterbi decoding :

Combine probabilities from the acoustic and language models to determine the overall most probable word sequence

Hello, I was just wondering what the guidance is for using AI to write code (especially in relation to the Speech Processing assignment, for example to parse through the data and write a text file with selections). The guidance about AI so far has been to do with the writing only so was wondering if you could clarify the rules for this kind of use?

Hello,
I am very interested in reading Atli’s Notion notes for advice about this assignment, but every time I try to access this page (even when logged into my university email) it says I need to request access. I have tried requesting access over a week ago, but every time I open the page it treats it as a new request I believe. Any advice?
Thanks

Is it possible to use tables or figures from a cited work, e.g. a proposed typology of a linguistic phenomenon, in our report? If so, is a screenshot permissible, or should we reproduce such tables or figures on our own?