I’m embarking on an exciting project to build a custom GPT model specifically tailored for translating into Taiwanese Mandarin. This initiative stems from the new feature by OpenAI, which allows individuals to create personalized GPT models without any coding skills.
Most of you might be aware that the current ChatGPT model is predominantly trained on text from Mainland China. However, Taiwanese Mandarin has its distinct nuances and variations. It’s not just about different words or phrases; it involves a unique blend of cultural and contextual elements that significantly influence the language.
To train this specialized GPT model, I require PDF files where this differences are explained.
If you have access to PDFs or any documents in Taiwanese Mandarin, especially those that highlight the linguistic distinctions from Mainland Mandarin, it would be immensely helpful.
The goal is to create a GPT model that can accurately understand and translate into Taiwanese Mandarin, respecting its unique cultural and linguistic attributes. When it’s done I will gladly share a link with everyone to use it. For now this in only available for premium subscribers of Open AI ChatGPT.
Yeah, this is good news. Taiwan needs AI that understands Taiwanese Mandarin. I don’t get it why open AI is ignoring it and I need to train own model. I also like to use text to speech AI for my business videos. There is hundred versions out there that speak Chinese with thick mainland china accent. Finally I found one that’s Taiwanese https://tts.yating.tw/
It’s not just OpenAI, and it’s not just translation software.
Most websites only give you Simplified and Traditional Chinese options, and do not make any distinctions between Hong Kong and Taiwan Traditional Chinese.
Because of this, Taiwanese people are already used to reading Hong Kong style (or generic) Traditional Chinese.
I’ve been using Claude to tell me about 台語 words used in Taiwan Mandarin. Glad I asked. Big difference between 可腦 and 機歪
I’d be very interested in the finished product. Still lots of literal translations and hallucinations in the bots I’ve used. Buggy.
I’d be interested in collaborating. I’m building a new transliteration tool that can take any Chinese text and convert it to modern Taiwanese Mandarin. It’s mostly not using AI, but I’m currently using openai for one-to-many/ambiguous character mappings from simplified to traditional since this is difficult to achieve without AI. Eventually, I plan to build my own model for this part of the process.
