I'm trying to figure out if building a voice-based sentiment analysis model for Brazilian portuguese would be possible. Since I couldn't find any free Brazilian datasets for voice analysis, would it be possible to develop the model using a dataset in english considering I won't need to transcribe the audio to text?
You could make it, but it probably wouldn’t work well. Portuguese and English have different phonology, vocabulary, prosody, and intonation. Those are the types of clues that your model would need to latch onto.
As with many ML problems on this site, it could be solved by collecting more data.