I used this list to create a dictionary and made it open source.

[erondpowell]

This isn't an issue just something I wanted to share.

I created a dictionary dataset out of this using AI last year. And just created a V2 with an expanded data set, so I made my V1 open source... I'll make v2 open source once I make v3. And plan to keep this cycle going.

This one was generated over a year ago, so it's not the complete wordlist as it is today - that can be fixed!

Lmk if you're interested in connecting on this!

https://github.com/freetalk-fun/freetalk-dictionary-v1

[thelabcat]

You lost me at "using AI".

[erondpowell]

I'm not sure what you mean? What's wrong with using AI?

Anyway, used it to make a Chrome browser extension. Still pretty MVP. But feel free to check it out:

https://chromewebstore.google.com/detail/freetalk-dictionary/hklbhgcopmfnjllpiheklhfjdgeiogig

🤠

[thelabcat]

Generative AI is very bad at getting the facts right. For word definitions, it mightn't be so bad since that has to do with the meaning of the words, which IS its designed purpose, but... Well, Gemma 3 thinks "gastroscopy" isn't a word. Did you double-check a large slew of the output definitions, just to make sure they were correct? Or, do I misunderstand what the deep AI was used for?

[erondpowell]

@thelabcat oh yeah! I got you. It took a lot of prompt engineering. There are still some issues with some of the terms. I plan to setup a way for folks to flag and re-prompt problem words.

Check out a few problem words if you want and lmk what you think.

[erondpowell]

Also, the v2 dataset - which I have not made public yet - is significantly better.

[Vasdranna]

Also, the v2 dataset - which I have not made public yet - is significantly better.

Will you happen to make it public, as well?

[erondpowell]

@Vasdranna Most likely, but not until I generate V3.

But take a look at V1 - it is very good, and has very few mistakes.
V2 is better in my opinion mainly because it provides more information.
Both have very, very few mistakes over all.
If you find any problems on V1, create an issue, and I can fix it.

Also, feel free to try this:
https://chromewebstore.google.com/search/FreeTalk%20Dictionary

[Vasdranna]

@Vasdranna Most likely, but not until I generate V3.

But take a look at V1 - it is very good, and has very few mistakes. V2 is better in my opinion mainly because it provides more information. Both have very, very few mistakes over all. If you find any problems on V1, create an issue, and I can fix it.

Also, feel free to try this: https://chromewebstore.google.com/search/FreeTalk%20Dictionary

Thank you!
Have you considered adding two or three example sentences for each entry?
I don't know if there's around another dictionary dataset as complete as yours.
Thanks for your work.

[erondpowell]

@Vasdranna yes! V2 has like 3-6 example sentences per distinct word meaning.