C-LARA

As a way to try out C-LARA’s content creation methods, I used English as the text language and annotations in French. For a prompt I asked for a poem to be generated based on the description I gave it of our new kitten Finley.

First point: the prompt should be saved and isn’t.

Second point: although you can ask for the text generated to be ‘improved’, as you give guidelines for that, it isn’t useful at the moment.

Because I wanted a more interactive experience, I opened up a conversation with Chat-4 in my own openai account and asked it for various versions of text about Finley, a Roger McGough style, a Shakespearean sonnet, and a Prc pic of pros with no lttr ‘e’. I cut and pasted these into my project. I like keeping a complete record of my interactions with Chat, so I am very happy to do this and cut and paste into C-LARA. It makes the process only seconds longer and you get a backup of the actual content, which is more important than the markups etc which can always be quickly recreated, admittedly at a small cost.

Then I went through the steps of segmenting, glossing and lemma tagging. Finally, I asked for a picture to be generated. I really didn’t like the picture and so I again went back to my own conversation with Chat-4 and generated a picture that way – since I could give it specific instructions about the picture, it’s not surprising that I got a result that I preferred. It was straightforward to download that to my computer and then upload it to my project. The resultant layout in the project needs to be fixed as the picture is too big.

There are lots of little things that can be improved and the odd bug that just this one project has uncovered – I hope that encourages others to try using it, as the more it is used, the easier it will be to discover what needs to be fixed.

I have to say I am amazed by how completely painless the process is.  Manny and Chat have created something that is a pleasure to use. It cost $1.36 and I should point out that this was for a relatively long amount of text compared with the short stories most likely to be generated at the moment. That said, I have done no proofreading/improvements and would not be capable of fixing the French if there are issues with it. My role was just to create something that would trouble-shoot the system and I seem to have succeeded. 🙂

You can find the project I created here

I’m on the sidelines of C-LARA, but I live with it all day everyday, and I’m thrilled to see such great progress being made, it’s very heartening! And exciting!!