Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

The use of emojis to determine sentiment is incredibly clever.

> The DeepThroat model is able to generate voices of varying degrees of emotion despite never having been exposed to emotive data of the character during training. Furthermore, multiple characters can be trained simultaneously, significantly reducing the amount of time required compared to if one were to train the character models individually.

Can't wait to read the paper.



That's interesting, I wonder if it can account for any drifts in meaning across generations: https://www.wsj.com/articles/sending-a-smiley-face-make-sure...


The emoji thing was a 4chan discovery. As was the idea to use both Arpabet and English, rather than just one of the two.




Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: