Are there drawbacks using ChatGPT for A2 Stories?

I’ve started learning European Portuguese two months ago and have worked through the mini-stories. I’ve been struggling to find in LingQ or on Youtube engaging comprehensible input for the next phase of reading and vocabulary building. I found that 3/4 of the content is either Brazilian Portuguese, a vastly different written and spoken dialect or is not appealing to me because it is childish stories, i.e. cooking recipes or stuff like “Anna goes to the supermarket” .

I’ve started using ChatGPT-4 to generate more advanced A2 mini-story style content in various tenses and points of view (1’st person, 3rd person).

I then take the ChatGPT-4 mini-stories and add a Narakeet generated audio and import it all into LingQ. The result is now I have plenty of interesting stories to read without spending endless hours searching Youtube, etc.

I was skeptical of ChatGPT-4 initially, but now that I’ve learned how to fine tune the prompting queries, I’am very impressed with what it can produce in seconds.

My question is are there any drawbacks to using ChatGPT to generate simple stories for this A2 vocabulary building phase after the LingQ mini-stories?

6 Likes

I don’t see a drawback. Maybe there’s a small chance there are errors here or there, but I personally wouldn’t worry about it. Especially as you fight through the beginning stages to get to higher level content that you may find more interesting.

3 Likes

As @Eric said, I don’t see any drawback either. I think it’s a great idea and you have a great strategy actually.

If you don’t mind sharing, I would be happy to read how you prompt on gpt-4, and your workflow.

I’m actually using more and more the integrated TTS feature on LingQ iOS, and it’s great anyway. More for laziness I admit, but it’s quite handy.

I often use these chatbots but I usually don’t import anything “yet”. Just for laziness as well. I wished they were integrated here but sooner or later they will, it’s just a question of time.

1 Like

My simplest workflow is using ForeFront ChatGpt (https://chat.forefront.ai) with a prompt like below.

“Write a story about the Portugal Carnation Revolution in European Portuguese (level A2) that is 10 sentences long. This story should be told twice. Once from the perspective of the main character, and once from a third party perspective.”

  1. I then correct any line breaks and paragraph breaks in a text editor and save the Portuguese output to a file (ie: PT - 2023-04-24 - Story About the Portuguese Carnation Revolution.txt)

This is optional but I also embed the filename and ChatGpt prompt I used in the first line of the text file like below. This between two empty [], this effectively creates a hidden comment in the text file with the info used to create it (my programmer training).

[]PT - P1 2023-04-24 - Story about the Portuguese Carnation Revolution

Write a story about the Portugal Carnation Revolution in European Portuguese (level A2) that is 10 sentences long. This story should be told twice. Once from the perspective of the main character, and once from a third party perspective.
[]

Eu sou a Ana, uma professora que vivia em Portugal durante a Revolução dos Cravos. Naquele tempo

  1. Setup Narakeet for the Portuguese voice and speed I like (this is done one time)

  2. Upload the output text above to NaraKeet

  3. Save the resulting NaraKeet m4a file to the same folder as I put the ChatGpt output text file, keeping the same file name with the m4a extension.

  4. Open the NaraKeet m4a file in audacity and re-export it with the same file name as and mp3 file with an .mp3 extension (LingQ Import requires mp3)

  5. Do a new LingQ Import Lesson (I already have a course setup that I assign to Mini Stories 2.0, this is optional.)

  6. In LingQ Import I select the ChatGpt text for the lesson

  7. In LingQ choose Audio File and select the mp3 file from NaraKeet

  8. Click Save & Generate Lesson and I’m done, the import is perfect and the audio is synced

The audio is synced perfectly in sentence mode and scrolling with the text.

FYI, I’ve had a native speaker check the stories grammar and NaraKeet pronunciation and she found them both near perfect.

3 Likes

Thanks for the workflow. I didn’t know about Forefront, is it something like Playground?

I’ve also checked NaraKeet, it seems they have a free and commercial account, or an option “à la carte”.

I just add a little money at a time to the NaraKeet account, like $10, it’s about $1 USD for 5 minutes of narration which I think is very reasonable.

FYI, I get roughly the same results using the free Microsoft Bing Chat (supposedly powered by ChatGpt4.

1 Like

FYI, I’ve attached an English translation of the the mini-story it created, in this case I asked for a historical story. For me the generated story, although A2 beginner material is quite substantive and interesting to read.

1 Like

You definitely can have a lot of fun with it. We have more and more interesting tools for learning languages nowadays.

1 Like