C-LARA

Here’s a summary of what’s been happening this week:

Melbourne student projects

We were lucky enough to be in Melbourne today when the students gave their final presentations. Things are going well. The students are keen to make sure that their work can be integrated into C-LARA, and it looks like we have a clear road-map for making sure that happens. In more detail:

1. Manual Text/Audio Alignment. Integration on this project is now almost complete.
2. Voice Recorder, Annotated Images. These projects look like they are almost done, and it should be easy to integrate them using Manual Text/Audio Alignment as a model.
3. Automatic Text/Audio Alignment. This should also be easy to integrate, but our impression was that the functionality is still quite fragile when there are mismatches or recognition errors. The question is whether the team will be able to improve it.
4. Flashcards. We had not previously been in contact with this team, but during the session we came up with a possible integration solution that seemed like it had fair chances of working.
5. Phonetic text. This is the project where it’s currently hardest to see a clear integration plan. We need to talk more with them.

All the projects have prepared good documentation and are putting their material in open source repositories so that we can use it freely. They will forward details to us soon. One of the first things we will do is show the new material to ChatGPT-4, to see how well it understands it.

Memory and plugins

We exchanged some messages with the provider of Papr Memory; it turned out that the latency problems we’d been experiencing were just a product of running it in a ChatGPT-4 thread that had got too long.

We have continued our discussion with the provider of the AskTheCode plugin, who has just deployed two new features we had discussed. We can’t yet use them, however, and have sent back a bug-report. He is very responsive and we expect the problems to be resolved soon. This plugin will be extremely useful for the handover of the Melbourne Uni student projects.

Conference presentations

Our paper for ALTA 2023 was accepted. The final version is due Oct 28.

We have a presentation at WorldCALL on Nov 9 and are discussing how to structure it.

UniSA server

We will not be able to do much on the Melbourne Uni student projects until mid-November, when they will have finished their exams, so for the next two weeks we will prioritise the UniSA server. We have a meeting scheduled for Friday with one of the sysadmins to discuss low-level issues.

Design

We said we would get back to thinking about design aspects of C-LARA when we had access to the visual-input version of ChatGPT-4. Now we do (or at least Cathy does). Another thing to discuss.

Next Zoom call

Note to Southern Hemisphere people: one hour later than it used to be due to Daylight Savings Time.

Thu Oct 26, 2023 19:00 Adelaide (= 08.30 Iceland = 09.30 Ireland/Faroe Islands = 10.30 Europe = 11.30 Israel = 12.00 Iran = 16.30 China = 19:30 Melbourne/New Caledonia)

Usual zoom link.