C-LARA

On Saturday a week ago, Lucreția and I (Claudia) used C-LARA to create 2 stories at the Romanian school Eindhoven. We had 2 classes: the younger kids (6-7 year olds) and older kids (I think 8-10 yo). Lucretia directed the activities and I used C-LARA. We projected the computer screen on the wall.

Next, I sent my feedback to the C-LARA team last week, and I’m also putting it here after talking to Lucreția, even though Manny addressed most of the items during the week.

We started with the younger kids and used the Simple C-LARA.

• Feedback 1. After the text is generated by ChatGPT, I’d like to be able to modify and save it. Right now I can only send it again to ChatGPT and it comes back with new text, close to the first one generated and modified, but still different.
• Feedback 2. The TTS split the Romanian word for penguin into 2 parts: when I hover over it I can hover over the first part of the word, or the second. It’s like hovering over 2 words, each with its own TTS. Since it was a story about a penguin, the word appeared more times in the story, in different forms, and still it was split. Strangely, I’ve just tried recreating the project using the same prompt and now it’s correct.
• Feedback 3. Definitely Simple C-LARA works better for what we’ve tried. When we used the normal C-LARA with the older kids, Lucretia did a great job at covering for the waiting times, but even then, when asking the kids for feedback, there was a complaint that it took too long. With the simple C-LARA, there was no complaint.
• Feedback 4. 2 of the younger kids came to the computer to press on the word to hear the TTS. One of them was my son, so maybe it doesn’t count, since he dares more with me, but there was also a girl and it was interesting for her too to trigger the speech for words and sentences. I think it’s more playing for them than intentional learning, and they like it when they do something that has an immediate consequence.

Normal C-LARA with the older kids

• Feedback 1. The tooltips help a lot, and it would be nice to also add the dependencies for the options that are disabled. For example, I wanted to have C-LARA get the mp3’s for the whole sentences, but the option (either Render Text or Render Phonetic Text – don’t remember exactly which one) was disabled and it took me a few minutes to find the part I still had to trigger. It would also be nice to have a way to know which steps were already triggered/done. Using the Simple C-LARA would have avoided this, but I wanted to show the older kids, besides the DALLE image, also an image we generated with the AI from Canva.
• Feedback 2. Because we wanted to involve the kids in thinking critically about the story generated by ChatGPT, we asked them to read it before we generated the rendered text. It was difficult for them to see the projected text. Later, because Lucreția also mentioned it, I finally remembered there’s an option to change the formatting of the text including font size, so we’ll use it next time.

General

The benefits of C-LARA are a bit different for younger kids than older ones. Older kids can intentionally work on their pronunciation and use the translation feature, since they’re more proficient in the translation language. You can also have a talk about AI, or point out particularities of the language. Younger kids don’t make a point about learning, they mostly follow instructions, so if you tell them to repeat the sentence they hear, they do. And are also playful.

We like C-LARA and want to use it further in the next weeks and months, and at the end of the school year present the stories to the parents during a special gathering. We’ll probably send the links to the stories before, so kids can practice at home if they want to. Lucretia and I are still brainstorming about the festivity, but it should include kids reading themselves, because developing reading abilities in Romanian is one of the main points for the classes.