Dictionary of AI Generated Example Sentences?

One problem with the dictionaries is the example sentences are too few in number and too communist. ChatGPT can generate fairly good example sentences. $20 worth of ChatGPT credit would be able to generate sentences for 10,000 words. I think having a dictionary containing generated example sentences would have a lot of value to users.
 

mikelove

皇帝
Staff member
Have you bought any add-on dictionaries? They add a whole lot more example sentences.

I wouldn't trust ChatGPT examples to be accurate enough to ship officially unless they were all hand-checked by a human, at which point the cost gets a lot more significant; I also don't trust it not to plagiarize an example sentence from a published dictionary, in which case that could create some complications doing business with that publisher, so its output would need to be checked for that too. So maybe we could make a "ChatGPT-assisted example sentence add-on," but it would have to be a collaboration between it and a human editor.

But our 4.0 update makes it very easy to make user-created example sentence databases, so if anybody feels like creating and sharing one of those that should be quite easy once 4.0 is out. It's possible - albeit a bit more awkward - even with our current app, e.g. with @Shun's massive "18,896 HSK Sentences" add-on https://plecoforums.com/threads/18-896-hsk-sentences.5615/.
 

Shun

状元
Hi Simon,

Nice idea! Just one question: Can ChatGPT also generate matching example sentences in English, French, German, and so on? (of nearly the same meaning?) If so, would you be willling to generate these sentence pairs with user donations and then upload them here? It would also be nice to auto-generate Text-to-Speech recordings of those sentences in Chinese and all other languages, so one could first hear the Chinese sentence and, after a delay, the same sentence in one of the other languages.

I've now also read Mike's post, of course he's right. As he has stated, if they are published by users, not by Pleco officially, the quality may be sufficient and copyright situation ought to be safe enough.

Best,

Shun
 
Yes, I have brought ABC, Tuttle and Guifan as well as all the free dictionaries. They help up to a point, but beyond a certain vocab level the example sentences are limited.

It sounds like 4.0 will be able to support example sentences in user dictionaries which will be great. I have created user dictionaries with generated example sentences but put them as part of the definition. That's okay but you don't get TTS.

Yes, ChatGPT can generate bilingual examples. I'll share my user dict and the files containing the generated sentences.
 

mikelove

皇帝
Staff member
That's correct, yes - you can both create them as example sentence entries to be part of the "Examples" tab and embed them in dictionary entries for the inset display / TTS / etc.
 

Shun

状元
Yes, ChatGPT can generate bilingual examples. I'll share my user dict and the files containing the generated sentences.
Thanks!

I would definitely be willing to chip in $10 in ChatGPT credit to generate more Chinese-English example sentence pairs. I'm sure you can also tell it of what difficulty the sentences should be. So perhaps you could make Beginner, Intermediate, Upper Intermediate, and Advanced pairs?

Can I send you an Apple Pay payment?

If one can get good TTS from Python, recording to an audio file, that could work great together with the ChatGPT sentence pairs for some 听力 exercises on the go.
 
Here is the dictionary import CSV file:


Download the dictionary-import-file, create a new user dictionary in Pleco and then import the file.

The words/ folder has the original example sentences from ChatGPT. The util/ folder has a couple of scripts for generating the sentences using ChatGPT and building the dictionary from the files in the words/ folder.

The difficulty of the sentences seems like normal spoken Chinese. I didn't play around with the prompt to try to generate different difficulties. The difficulty seems right for the vocab level.

As Pleco already has TTS I'm not intending to do generate sound files. We will have to wait for 4.0 to be able to get it though.
 

Shun

状元
It looks wonderful, thanks! Yes, the sentence difficulty seems to be adequate. It's quite amazing to see what ChatGPT can do here. Humans wouldn't have the patience to come up with all of these.

Yeah, sorry about the misunderstanding, I meant the TTS remark as something for myself to program.

If you wish, you could also upload the dictionary file directly to the Pleco forums; that way probably even more users would download it.
 
Top