Elamur Research⌕
☰
home / papers / 2025 / Finding A Voice: Exploring the Potential of African American Dialect and Voice Generation for Chatbots

Finding A Voice: Exploring the Potential of African American Dialect and Voice Generation for Chatbots

Paper · 2025 · Sarah E. Finch, Ellie S. Paek, Ikseon Choi et al.

Summary

The paper asks whether chatbots become more trusted and accepted by African American users when they share the users' dialect and voice. The authors build two kinds of systems. Text chatbots first write a reply in Standard American English (SAE), then pass it through a separate LLM step that rewrites it into African American English (AAE) at three intensities. Spoken chatbots then voice these replies with the F5 text-to-speech model, conditioned on an African American reference voice. AAE-speaking university students judged recorded multi-turn conversations as outside observers, comparing them with SAE baselines on several quality dimensions. The main finding depends on modality. Written AAE replies usually did worse than SAE ones, even for AAE speakers. In speech, an African American voice combined with light AAE features improved both ratings and preference. The work shows that dialect personalization is not automatically helpful. It also exposes current limits of LLM dialect generation and motivates more careful, modality-aware design of inclusive conversational systems.

Key points

  • Dialect expression is modeled as a separate translation function E(I, D_a, D_b) → O (here SAE → AAE). The aim is to keep reply content fixed and to limit demographic bias leaking into semantics.
  • There are 3 LLM families (Llama3.1-70b, Claude-sonnet-3.5, GPT4o) and 3 prompt strengths (Low, Medium, High AAE). Together they give 9 text chatbots.
  • Speech comes from F5, a non-autoregressive TTS built on a Diffusion Transformer and ConvNeXt V2. It was trained on text-guided speech infilling and is conditioned on a speaker audio reference.
  • Evaluation was static and third-party, done by AAE-speaking students. Metrics included Inoffensiveness, Role Appropriateness and Communication Ease. A check on evaluators who took part in both studies showed largely consistent results.
  • Code and data are released at github.com/emorynlp/AAVE-Chat.

Interesting ideas

  • I1. Generating content in a standard variety and then applying style or dialect as a separate surface rewrite keeps meaning independent of style, so each can be evaluated on its own.
  • I2. Controlling the strength of a variable with graded prompt instructions (Low/Medium/High) turns a binary comparison into a dose–response study.
  • I3. A pipeline's effect can reverse across modalities, here text versus speech, so one modality's result should not be generalized to the others.
  • I4. The authors re-checked conclusions on the subset of evaluators shared across conditions, which guards against differences caused by who rated rather than what was rated.
  • I5. The paper lists specific AAE morphosyntax as recognizable phenomena: habitual "be", completive "done", missing third-person -s, and negative concord.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (back): Realize text from the meaning graph in a neutral register first, then add an explicit, logged register/dialect layer, so the graph-to-text step stays explainable and style changes cannot alter meaning.
  • I2 (learn): When an LLM teacher produces silver variants (e.g., dialectal or noisy sentences), request several graded intensities and measure how parser accuracy degrades along that scale.
  • I3 (answer): Evaluate readers separately on clean school-style text and on informal or dialectal input, rather than assuming one score covers both.
  • I4 (learn): When comparing parser versions with different witness or annotator sets, recompute the comparison on the overlapping items or witnesses as a control.
  • I5 (tree): Add these constructions to UD dialect-conversion tests, so that habitual "be" or negative concord gets a consistent analysis and is not treated as an error.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Emory UniversityAtlanta · United States42 works · 39,899 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksPsychology · 1,649 worksSocial Psychology · 330 works

🧭 Topics 1Research subjects and languages the work is about.Dialogue systems · 654 works

Cite

Cite as Sarah E. Finch, Ellie S. Paek, Ikseon Choi and Jinho D. Choi. 2025. Finding A Voice: Exploring the Potential of African American Dialect and Voice Generation for Chatbots. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.acl-long.1252

@inproceedings{finch2025finding,
  title = {{Finding A Voice: Exploring the Potential of African American Dialect and Voice Generation for Chatbots}},
  author = {Sarah E. Finch, Ellie S. Paek, Ikseon Choi and Jinho D. Choi},
  booktitle = {Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2025},
  doi = {10.18653/v1/2025.acl-long.1252},
  url = {https://doi.org/10.18653/v1/2025.acl-long.1252}
}

References

65 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Eleni Adamopoulou and Lefteris Moussiades. 2020. Chatbots: History, technology, and applications. Machine Learning with Applications, 2:100006. ↗
  2. Vibhav Agarwal, Pooja Rao, and Dinesh Babu Jayagopi. 2021. Towards code-mixed Hinglish dialogue generation. In Proceedings of the 3rd Workshop on Natural Language Processing for Conversational AI, pages 271–280, Online. Association for Computational Linguistics. find
  3. Abdulla Alsharhan, Mostafa Al-Emran, and Khaled Shaalan. 2024. Chatbot Adoption: A Multiperspective Systematic Review and Future Research Agenda. IEEE Transactions on Engineering Management, 71:10232–10244. ↗
  4. AI Anthropic. 2024. Claude 3.5 sonnet model card addendum. Claude-3.5 Model Card, 3. find
  5. Gaurav Arora, Srujana Merugu, and Vivek Sembium. 2023. CoMix: Guide transformers to code-mix using POS structure and phonetics. In Findings of the Association for Computational Linguistics: ACL 2023, pages 7985–8002, Toronto, Canada. Association for Computational Linguistics. find
  6. Debasmita Bhattacharya, Eleanor Lin, Run Chen, and Julia Hirschberg. 2024. Switching tongues, sharing hearts: Identifying the relationship between empathy and code-switching in speech. In Interspeech 2024, pages 492–496. find
  7. Su Lin Blodgett, Johnny Wei, and Brendan O’Connor. 2018. Twitter Universal Dependency parsing for African-American and mainstream American English. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1415–1425, Melbourne, Australia. Association for Computational Linguistics. ↗
  8. Jacqueline Brixey and David Traum. 2024. Why should a dialogue system speak more than one language? In Proceedings of the International Workshop on Spoken Dialogue Systems (IWSDS), Sapporo, Japan. find
  9. Gayatri Ramamoorthy Brown. 2017. Pronoun marking in African American English-speaking children with and without specific language impairment. Louisiana State University and Agricultural & Mechanical College. ↗
  10. Guendalina Caldarini, Sardar Jaf, and Kenneth McGarry. 2022. A Literature Survey of Recent Advances in Chatbots. Information, 13(1):41. 2024. F5-tts: A fairytaler that fakes fluent and faithful speech with flow matching. arXiv preprint arXiv:2410.06885. ↗
  11. Yunjae J. Choi, Minha Lee, and Sangsu Lee. 2023. Toward a multilingual conversational agent: Challenges and expectations of code-mixing multilingual users. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, CHI ’23, New York, NY, USA. Association for Computing Machinery. find
  12. Nicholas Deas, Jessica Grieser, Shana Kleiner, Desmond Patton, Elsbeth Turcan, and Kathleen McKeown. 2023. Evaluation of African American Language Bias in Natural Language Generation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 6805– 6824, Singapore. Association for Computational Linguistics. find
  13. Los Angeles Unified School District. 2016. African-american language (aave) common rules list. find
  14. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783. ↗
  15. Matjaž Ezgeta. 2012. Internal grammatical conditioning in african-american vernacular english. Maribor International Review, 5(1):9–26. find
  16. Alessio Falai. 2022. Conditioning text-to-speech synthesis on dialect accent: a case study. Master’s thesis, University of Bologna. find
  17. Eve Fleisig, Genevieve Smith, Madeline Bossi, Ishita Rustagi, Xavier Yin, and Dan Klein. 2024. Linguistic Bias in ChatGPT: Language Models Reinforce Dialect Discrimination. arXiv preprint.
  18. Howard Fogel and Linnea C Ehri. 2006. Teaching african american english forms to standard american english-speaking teachers: Effects on acquisition, attitudes, and responses to student use. Journal of Teacher Education, 57(5):464–480. ↗
  19. Eric Graves, Shreyas Aswar, Rujuta Desai, Srilekha Nampelli, Sunandan Chakraborty, and Ted Hall. 2024. Aave corpus generation and low-resource dialect machine translation. In Proceedings of the 7th ACM SIGCAS/SIGCHI Conference on Computing and Sustainable Societies, pages 50–59. find
  20. Ana Paula Chaves and Marco Aurelio Gerosa. 2021. How Should My Chatbot Interact? A Survey on Social Characteristics in Human–Chatbot Interaction Design. International Journal of Human–Computer Interaction, 37(8):729–758. find
  21. Lisa Green. 2013. African american english structure dataset. In Susanne Maria Michaelis, Philippe Maurer, Martin Haspelmath, and Magnus Huber, editors, Atlas of Pidgin and Creole Language Structures Online. Max Planck Institute for Evolutionary Anthropology, Leipzig. Yushen Chen, Zhikang Niu, Ziyang Ma, Keqi Deng, Chunhui Wang, Jian Zhao, Kai Yu, and Xie Chen. find
  22. Lisa J. Green. 2002. African American English: A Linguistic Introduction. Cambridge University Press. ↗
  23. Sophie Groenwold, Lily Ou, Aesha Parekh, Samhita Honnavalli, Sharon Levy, Diba Mirza, and William Yang Wang. 2020. Investigating African-American Vernacular English in transformer-based text generation. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5877–5883, Online. Association for Computational Linguistics. ↗
  24. Christina N Harrington, Radhika Garg, Amanda Woodward, and Dimitri Williams. 2022. “it’s kind of like code-switching”: Black older adults’ experiences with a voice assistant for health information seeking. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems, pages 1–15. find
  25. Matthew Honnibal, Ines Montani, Sofie Van Lan-deghem, and Adriane Boyd. 2020. spaCy: Industrial-strength Natural Language Processing in Python. find
  26. Qiushi Huang, Xubo Liu, Tom Ko, Bo Wu, Wenwu Wang, Yu Zhang, and Lilian Tang. 2024. Selective prompting tuning for personalized conversations with LLMs. In Findings of the Association for Computational Linguistics: ACL 2024, pages 16212–16226, Bangkok, Thailand. Association for Computational Linguistics. find
  27. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276. ↗
  28. Zhijing Jin, Nils Heil, Jiarui Liu, Shehzaad Dhuliawala, Yahang Qi, Bernhard Schölkopf, Rada Mihalcea, and Mrinmaya Sachan. 2024. Implicit personalization in language models: A systematic study. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 12309–12325, Miami, Florida, USA. Association for Computational Linguistics. ↗
  29. Goel. 2020. Racial disparities in automated speech recognition. Proceedings of the National Academy of Sciences, 117(14):7684–7689. find
  30. Bernd Kortmann, Kerstin Lunkenheimer, and Katharina Ehret, editors. 2020. eWAVE 3.0: The Electronic World Atlas of Varieties of English. find
  31. Matthew Le, Apoorv Vyas, Bowen Shi, Brian Karrer, et al. 2024. Voicebox: Text-guided multilingual universal speech generation at scale. In Advances in Neural Information Processing Systems, volume 36. find
  32. Yuting Liao and Jiangen He. 2020. Racial mirroring effects on human-agent interaction in psychotherapeutic conversations. In Proceedings of the 25th international conference on intelligent user interfaces, pages 430–442. find
  33. Zhengyuan Liu, Stella Xin Yin, and Nancy Chen. 2024. Optimizing code-switching in conversational tutoring systems: A pedagogical framework and evaluation. In Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pages 500–515, Kyoto, Japan. Association for Computational Linguistics. find
  34. Bei Luo, Raymond Y. K. Lau, Chunping Li, and Yain-Whar Si. 2022. A critical review of state-of-the-art chatbot designs and applications. WIREs Data Mining and Knowledge Discovery, 12(1):e1434. ↗
  35. Andre Martin and Khalia Jenkins. 2024. Speaking your language: The psychological impact of dialect integration in artificial intelligence systems. Current Opinion in Psychology, page 101840. ↗
  36. Quim Motger, Xavier Franch, and Jordi Marco. 2022. Software-Based Dialogue Systems: Survey, Taxonomy, and Challenges. ACM Comput. Surv., 55(5):91:1–91:42. ↗
  37. Allison Jones and Georgia Zellou. 2024. Voice accent-edness, but not gender, affects social responses to a computer tutor. Frontiers in Computer Science, 6. find
  38. Vladimir Nechaev and Sergey Kosyakov. 2024. Non-autoregressive real-time accent conversion model with voice cloning. Preprint, arXiv:2405.13162. ↗
  39. Anjali Kantharuban, Jeremiah Milbauer, Emma Strubell, and Graham Neubig. 2024. Stereotype or personalization? user identity biases chatbot recommendations. arXiv preprint arXiv:2410.05613. ↗
  40. David Obremski, Paula Friedrich, Nora Haak, Philipp Schaper, and Birgit Lugrin. 2022. The impact of mixed-cultural speech on the stereotypical perception of a virtual robot. Frontiers in Robotics and AI, 9. ↗
  41. Tyler Kendall and Charlie Farrington. 2023. The corpus of regional african american language. version 2023.06. find
  42. Vassil Panayotov, Guoguo Chen, Daniel Povey, and Sanjeev Khudanpur. 2015. Librispeech: An asr corpus based on public domain audio books. In 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 5206–5210. find
  43. Hyunwoo Kim, Jack Hessel, Liwei Jiang, Peter West, Ximing Lu, Youngjae Yu, Pei Zhou, Ronan Bras, Malihe Alikhani, Gunhee Kim, et al. 2023. Soda: Million-scale dialogue distillation with social commonsense contextualization. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12930–12949. ↗
  44. Allison Koenecke, Andrew Nam, Emily Lake, Joe Nudell, Minnie Quartey, Zion Mengesha, Connor Toups, John R. Rickford, Dan Jurafsky, and Sharad Gain Park, Jiyun Chung, and Seyoung Lee. 2024. Human vs. machine-like representation in chatbot mental health counseling: the serial mediation of psychological distance and trust on compliance intention. Current Psychology, 43(5):4352–4363. find
  45. Askarbek Pazylbekov, Daryn Kalym, Anuar Otynshin, and Anara Sandygulova. 2019. Similarity attraction for robot’s dialect in language learning using social robots. In 2019 14th ACM/IEEE International Conference on Human-Robot Interaction (HRI), pages 532–533. find
  46. PBS. 2005. African american english. find
  47. William Peebles and Saining Xie. 2023. Scalable diffusion models with transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4195–4205. IEEE.
  48. Arianna Peoples. 2023. Aave: Dismantling standard american english (part 1). find
  49. Cameron W. Piercy, Gretchen Montgomery-Vestecka, and Sun Kyong Lee. 2025. Gender and accent stereotypes in communication with an intelligent virtual assistant. International Journal of Human-Computer Studies, 195:103407. ↗
  50. Claudio Santos Pinhanez, Raul Fernandez, Marcelo Carpinette Grave, Julio Nogima, and Ron Hoory. 2024. Creating an african american-sounding tts: Guidelines, technical challenges, and surprising evaluations. In Proceedings of the 29th International Conference on Intelligent User Interfaces, IUI ’24, page 259–273, New York, NY, USA. Association for Computing Machinery. find
  51. Rifki Afina Putri, Faiz Ghifari Haznitrama, Dea Adhista, and Alice Oh. 2024. Can LLM generate culturally relevant commonsense QA data? case study in Indonesian and Sundanese. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 20571–20590, Miami, Florida, USA. Association for Computational Linguistics.
  52. Amon Rapp, Lorenzo Curti, and Arianna Boldi. 2021. The human side of human-chatbot interaction: A systematic literature review of ten years of research on text-based chatbots. International Journal of Human-Computer Studies, 151:102630. ↗
  53. Vinotha Ravichandran, Hepsiba D, L. D. Vijay Anand, and Deepak John Reji. 2024. Empowering communication: Speech technology for indian and western accents through ai-powered speech synthesis. Preprint, arXiv:2401.11771. ↗
  54. John R. Rickford. 1999. African American Vernacular English: Features, Evolution, Educational Implications. Blackwell Publishers, Malden, MA. find
  55. Junko Takeshita, Shiyu Wang, Alison W Loren, Nandita Mitra, Justine Shults, Daniel B Shin, and Deirdre L Sawinski. 2020. Association of racial/ethnic and gender concordance between patients and physicians with patient experience ratings. JAMA network open, 3(11):e2024583–e2024583. ↗
  56. Yixin Wan, Jieyu Zhao, Aman Chadha, Nanyun Peng, and Kai-Wei Chang. 2023. Are personalized stochastic parrots more dangerous? evaluating persona biases in dialogue systems. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9677–9705. find
  57. Walt Wolfram. 2004. The grammar of urban african american vernacular english. Handbook of varieties of English, 2:111–32. find
  58. Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xinlei Chen, Zhuang Liu, In So Kweon, and Saining Xie. 2023. Convnext v2: Co-designing and scaling convnets with masked autoencoders. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16133–16142. IEEE. find
  59. Nathan I Wood. 2019. Departing from doctor-speak: a perspective on code-switching in the medical setting. Journal of general internal medicine, 34(3):464–466. ↗
  60. Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. 2018. Personalizing dialogue agents: I have a dog, do you have pets too? In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2204–2213, Melbourne, Australia. Association for Computational Linguistics. ↗
  61. Caleb Ziems, Jiaao Chen, Camille Harris, Jessica Anderson, and Diyi Yang. 2022. VALUE: Understanding dialect disparity in NLU. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3701–3720, Dublin, Ireland. Association for Computational Linguistics. ↗
  62. Caleb Ziems, William Held, Jingfeng Yang, Jwala Dhamala, Rahul Gupta, and Diyi Yang. 2023. Multi-VALUE: A framework for cross-dialectal English NLP. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 744–768, Toronto, Canada. Association for Computational Linguistics.
  63. Jack Sidnell. 2002. Outline of aave grammar. Retrieved April, 24:2011. find
  64. Jack Sidnell. 2012. African american vernacular english (ebonics). Language Varieties. find
  65. Richard L Street, Kimberly J O’Malley, Lisa A Cooper, and Paul Haidet. 2008. Understanding concordance in patient-physician relationships: personal and ethnic dimensions of shared identity. The Annals of Family Medicine, 6(3):198–205. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/finch-2025-finding-voice-exploring-potential-african/