C-LARA

I thought it might be useful to post a summary every Wednesday of what’s been going on during the preceding week, so that people can take a look at it before the Thursday Zoom call. Here’s a first try. Feedback on how to adjust the format will be appreciated!

Melbourne student projects

Three of the Melbourne student projects have reached the point where we can start concrete work on integrating them into C-LARA:

1. Voice Recorder. A tool that makes it possible to record human audio to integrate into a C-LARA document as an alternative to TTS, when people want to create a high-quality document. The design is similar to the “LiteDevTools” recorder than was used in LARA.
2. Manual Text/Audio Alignment. A different version of the above: again, we want to use human audio, but this time the assumption is that we have a single mp3 already available. The tool makes it easy to add annotations showing how to cut this up into pieces that match the defined segments.
3. Annotated images. Here, the idea is to make it possible to associate text with an image and provide a tool which lets the user mark the area associated with each word. Again, we had similar functionality in LARA, which we for example used in the alphabet book project.

I have added initial views in the Heroku deployment to support these functionalities. In preparation for integrating with the actual student projects, I have been testing the first two with LiteDevTools and Audacity, uploading and downloading files by hand. The third one still has only very basic support for images. I am currently working with Chat on improving this.

The projects are supposed to finish on Oct 25, though there is a possibility of extending this by up to three weeks if the students want to do so.

Memory and plugins

After the interesting discussion last week on memory, I looked around and found the Papr Memory plugin, which already provides ChatGPT-4 with the kind of external memory we talked about. You can install it directly from the plugin store. So far, it works well: Chat seems to have much better global understanding of what we’re doing.

Chat and I have also been experimenting with AskTheCode, a plugin that lets the AI read a GitHub repository. We had some issues with this and contacted the plugin’s creator, who was extremely helpful both in explaining the nature of the problems and what could be done to address them. If he implements the enhancements we talked about, and it sounded like he would, this will also be tremendously useful.

Multi-Word Expressions

I started an Overleaf document which summarises discussions so far. Let’s add to it and start doing some concrete experimentation!

Next Zoom call

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

Thu Oct 12, 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.