Accent Training with AI: Promise and Pitfalls
A central theme of the discussion revolves around the potential of AI-powered accent training tools like BoldVoice for language learners. Many users express excitement about "real-time accent feedback," a capability that "language learners have never had throughout all of human history, until now" (georgewsinger). The ability to "show a language learner what they would sound like with a more native accent" (adhsu01) is seen as a significant advantage.
However, some users, like treetalker, noted discrepancies between the visual representation of progress and actual improvements in intelligibility: "That said, I found the recording of Victor's speech after practicing with the recording of his own unaccented voice to be far less intelligible than his original recording." JoshTko echoed this sentiment, stating, "Visually I'd say victor improved by at most 5% and not 50% as indicated by the visualization. In some regards it was even harder to understand than the original due to speed and cadence without improvement in core pronunciation."
The Nuances of "Neutral" Accents and Intelligibility
The discussion delves into the complexities of accents, moving beyond a simple "native vs. non-native" dichotomy. pjc50 highlights the "idea that accents are a complex statistical distribution" and cautions against the concept of a "default" or "neutral" accent. They point out that "There's always the tendency for people to say 'my accent is the neutral standard against which all others should be measured'."
ilyausorov of BoldVoice acknowledges this nuance, stating, "I don't think we ever use the term default or neutral. The 'the American English accent of our expert accent coach Eliza' is just that -- it's one accent." They further clarify that while a pedagogical direction is necessary, "we do need to set some kind of direction in our pedagogy, but we 100% recognize that there isn't just 1 American English accent, and there's lots of variance."
asveikau raises an important point about intelligibility, arguing, "IMO it's perfectly ok to have an accent, as long as the speech meets some baseline of intelligibility." They provide a specific example: "Victor's problem isn't really the vowels or pacing. The final consonants are soft or not really audible."
Accessibility and Alternatives to Traditional Coaching
The expense and availability of traditional accent coaches is also a recurring theme. anadalakra notes that "Speech coaches are absolutely the way to go, but they're outside the price range for most people ($200+/hr for a good coach)." They position BoldVoice as providing "coach-level feedback and instruction at a price point that everyone can access, on demand."
pjc50 counters the assertion that real-time accent feedback is entirely novel, stating: ".. unless they had access to a native speaker and/or vocal coach? While an automated Henry Higgins is nifty, it's not something humans haven't been able to do themselves."
Ethical Concerns and Data Privacy
Several users express concerns about the privacy implications of using such services. fxtentacle initially felt "excited" but then reconsidered after reading the privacy policy: "They want permission to save all of my audio interactions for all eternity. It's so sad that I will never try out their (admittedly super cool) AI tech."
While anadalakra clarifies that users can request data deletion, fxtentacle points out the limitations of opting out: "Yeah, I can opt out. By not using any voice-related feature in their voice training app." This exchange highlights the tension between personalized feedback and data retention policies.
Potential Applications Beyond Accent Reduction
Several participants suggest alternative or expanded applications for the technology. vessenes proposes exploring "American idiolects," specifically mentioning AAVE and Gay male speech. They express disappointment with ChatGPT's handling of these sociolects, noting, "It seems to have a list of 'okay' and 'bad' idiolects baked in." They envision an "idiolect-manager, something that could help me move my speech more or less toward a given idiolect." This suggestion points to the potential for using the technology to understand and even emulate different dialects and speech patterns beyond simply minimizing a foreign accent. vessenes also points to specific market opportunities, such as "Voice coaches in Hollywood."
Model Perception and Influencing Factors
wbroo poses an important question about the factors influencing the model's perception: "Have you tested for other factors like speaking speed, emotional tone, or microphone quality to see what else is (or isn’t) influencing model perception?" This highlights the need for researchers to consider a range of variables when evaluating the effectiveness of accent training AI.