C-LARA

Phonetic texts

A first version of the phonetic text functionality is now more or less complete. Here are the main points:

• We support two versions: regular letter-to-sound rules, based on a conversion table, and complex/irregular letter-to-sound rules, based on a phonetic lexicon. For languages it knows well enough, GPT-4 can in the second scheme automatically add missing phonetic lexicon entries.
  You can also combine the two schemes, in which case the phonetic lexicon is treated as providing a list of exceptions to the regular rules. I’ve been testing the “regular” version using the New Caledonian language Drehu, and the “irregular” version using English and French.
• All the information used can be uploaded and edited through the C-LARA interface. So you can enter/edit letter-to-sound rules, upload phonetic lexica, and review/edit phonetic entries and alignments guessed by C-LARA.
• You can upload audio files to attach to the phonetic entries.
• Phonetic texts can be rendered in the same way as normal ones, to create multimedia texts. If you create both normal and phonetic versions of a project, each page links to the corresponding one in the other version.

Images

I have cleaned up the interface we use for handling images. It’s now simple to upload and revise images attached to a C-LARA text.

Alphabet book for Drehu

Pauline and I have been testing these new capabilities in the context of an alphabet book for Drehu, which we’ve been putting together over the last few days. The content is based on an existing paper version created by the Académie des Langues Kanak, but I am creating new images using DALL-E-3. It’s now about 30% complete. The draft version is posted at

https://c-lara.unisa.edu.au/accounts/rendered_texts/17/normal/page_1.html

So far no audio, but Pauline says she will send it soon.

Easy creation of C-LARA picture-books

Putting together the alphabet book is fun, as we found on our previous LARA alphabet book project, but it’s still rather laborious. I think it should be fairly easy to adapt the new image uploading screen, which looks like this,

to provide an interface for creating simple picture-books, where each page contains an image and a sentence or two of text. The layout would be more or less the same, with an image, an upload control, an image name, a page number, and a delete control, but we would add a box for the text. The backend processing involved should be straightforward.

Upcoming conference submissions

Chat and I submitted an abstract about phonetic texts to INTED. Pauline and I are planning to submit a paper about the Drehu work to ComputEL-7, due Dec 15.

Password reset and email updates

Many people have asked for a password reset function. As a stopgap, I added a manual password reset that people with admin privileges can use – so if you’ve lost your password, mail me and I will reset it for you.

We heard today that the email address we requested for the UniSA server is now ready for use. I should be able to use it to add a proper password reset function, also social network updates.

Next Zoom call

Note to Southern Hemisphere people: two hours later than winter time, i.e. same time as last week.

Thu Dec 14, 2023, 20:00 Adelaide (= 09.30 Iceland = 09.30 Ireland/Faroe Islands = 10.30 Europe = 11.30 Israel = 13.00 Iran = 17.30 China = 20:30 Melbourne/New Caledonia)