Weekly summary: Nov 17 – Nov 22, 2023

Here is this week’s roundup:

UniSA server

The UniSA server appears to be fully operational. We have now been running it for a couple of weeks, no one has reported any issues.

GPT-4 Turbo

I reran the Swedish glossing operation for the six English examples in our upcoming ALTA paper. Here are the error rates:

As you can see, GPT-4 Turbo seems to be a lot better. The outlier was the CH (children’s story) task: for unobvious reasons, a stretch of about 25 words was not glossed at all. We have often seen this kind of random variation.

Melbourne student projects

We have now received full reports and code for all six projects. I am reviewing this to see what needs to be done to complete the integration.

“Phonetic texts”

Pauline Welby (CNRS/University of New Caledonia) is visiting this week; we have a common interest in “phonetic texts”, texts which are annotated at the letter level so that a student can click on letter groups and get pronunciation information. We worked on this idea last year under LARA (there is a paper here), and I’ve now ported over some of the functionality to C-LARA. So far, it only handles the simple case where the language is spelled completely phonetically. The LARA code can deal quite well with languages like English and French which have complex letter/sound correspondences, and I will later move over that functionality too.

General improvements

At last weeks’s Zoom meeting, Belinda, Rina and I put together a list of general improvements we should plan to add soon to the platform:

Search functionality for registered texts and own projects.
Export and import of projects in zipfile form.
Password retrieval.
Automatic payment integration.
Online help.
Email notifications for social network.
Following and friending.
CSS styling and other basic design features.

I have added (1) and started working on (2).

(3) and (6) require integration of a mail server – this is currently being discussed with the UniSA sysadmins, who are working on setting up a C-LARA mail address.

Discord server

Jeremiah McPadden has set up a Discord server for C-LARA – you will find it here. We should try using this for technical discussions, feature requests and similar.

Next Zoom call

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

Thu Nov 23, 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)