Rabbit-AAC: Uma Arquitetura Híbrida para Recomendação de Texto em Comunicação Aumentativa e Alternativa

• Fernando Torres Ferreira da Silva USP
• Bruna C. R. Cunha USP

DOI: https://doi.org/10.5753/webmedia.2026.30554

Resumo

Conditions such as Cerebral Palsy (CP) and Amyotrophic Lateral Sclerosis (ALS) severely impair oral communication, making Augmentative and Alternative Communication (AAC) essential for supporting user autonomy. However, the low typing speed of virtual keyboards, particularly when using indirect access methods such as eye-tracking, remains a major limitation and contributes to high rates of AAC technology abandonment. To address this challenge, we propose and evaluate Rabbit-AAC, a hybrid architecture for text recommendation in AAC systems. The proposed approach integrates three complementary strategies: (i) a low-latency autocomplete mechanism based on a Trie data structure and a frequency-based lexical database to accelerate intra-word typing; (ii) an optimized reranking module leveraging different categories of Large Language Models (LLMs); and (iii) parameter-efficient fine-tuning using Low-Rank Adaptation (LoRA) on the two most efficient models. Extensive experiments on the synthetic AAC conversation dataset (willwade/AACConversations) show that the BERTimbau + LoRA + Trie architecture achieves a Mean Reciprocal Rank (MRR) of 0.730, Top-5 accuracy of 81.1%, and an average latency of 51 ms, providing the best balance between semantic accuracy and real-time performance. In contrast, the multitoken reranking strategy based on GPT-2 + LoRA achieves a Top-5 accuracy of 99% but incurs a prohibitive average latency of 1.28 s, highlighting the importance of carefully balancing effectiveness and responsiveness when designing AI-powered AAC systems. These results demonstrate that combining classical data sctructures with parameter-efficient fine-tuning offers a viable path toward real-time, contextaware AAC systems.

Palavras-chave: augmentative and alternative communication, text prediction, hybrid architecture, bertimbau, lora, trie, reranking, assistive technology, natural language processing, eye-tracking

Referências

Dallin J Bailey, Francesca Herget, Derek Hansen, Forrest Burton, Grant Pitt, Tyson Harmon, and David Wingate. 2026. Generative AI applied to AAC for aphasia: a pilot study of aphasia-GPT. Aphasiology 40, 1 (2026), 150–165.

Jean-Dominique Bauby. 1998. The Diving Bell and the Butterfly. Random House, New York.

Wenxi Chen, Ziyang Ma, Xiquan Li, Xuenan Xu, Yuzhe Liang, Zhisheng Zheng, Kai Yu, and Xie Chen. 2025. Slam-AAC: Enhancing audio captioning with paraphrasing augmentation and clap-refine through LLMs. In ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 1–5.

Michael Collins and Terry Koo. 2005. Discriminative Reranking for Natural Language Parsing. Computational Linguistics 31, 1 (2005), 25–70. DOI: 10.1162/0891201053630273

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of NAACLHLT 2019. Association for Computational Linguistics, 4171–4186. DOI: 10.18653/v1/N19-1423

Alexander Fiannaca, Ann Paradiso, Mira Shah, and Meredith Ringel Morris. 2017. AACrobat: Using Mobile Devices to Lower Communication Barriers and Provide Autonomy with Gaze-Based AAC. In Proceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing (Portland, Oregon, USA) (CSCW ’17). Association for Computing Machinery, New York, NY, USA, 683–695. DOI: 10.1145/2998181.2998215

Paul M. Fitts. 1954. The Information Capacity of the Human Motor System in Controlling the Amplitude of Movement. Journal of Experimental Psychology 47, 6 (1954), 381–391. DOI: 10.1037/h0055392

Tom Griffiths, Steven Bloch, Katie Price, and Michael Clarke. 2019. 7 - Alternative and Augmentative Communication. In Handbook of Electronic Assistive Technology, Donna Cowan and Ladan Najafi (Eds.). Academic Press, 181–213. DOI: 10.1016/B978-0-12-812487-1.00007-7

Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, and Weizhu Chen. 2021. LoRA: Low-Rank Adaptation of Large Language Models. CoRR abs/2106.09685 (2021). arXiv:2106.09685 [link]

Macarious Hui, Jinda Zhang, and Aanchan Mohan. 2025. Enhancing AAC software for dysarthric speakers in e-health settings: an evaluation using torgo. In ICC 2025-IEEE International Conference on Communications. IEEE, 3673–3679.

Shaun K. Kane and Meredith Ringel Morris. 2017. Let’s Talk About X: Combining Image Recognition and Eye Gaze to Support Conversation for People with ALS. In Proceedings of the 2017 Conference on Designing Interactive Systems (Edinburgh, United Kingdom) (DIS ’17). Association for Computing Machinery, New York, NY, USA, 129–134. DOI: 10.1145/3064663.3064762

I. Scott MacKenzie. 1992. Fitts’ Law as a Research and Design Tool in Human-Computer Interaction. Human-Computer Interaction 7, 1 (1992), 91–139. DOI: 10.1207/s15327051hci0701_3

Sharynne McLeod. 2018. Communication rights: Fundamental human rights for all. International journal of speech-language pathology 20, 1 (2018), 3–11.

Darius Miniotas. 2000. Application of Fitts’ Law to Eye Gaze Interaction. In CHI ’00 Extended Abstracts on Human Factors in Computing Systems. ACM, 339–340. DOI: 10.1145/633292.633496

Jakob Nielsen et al. 1993. Usability engineering. Vol. 1. Morgan Kaufmann San Francisco.

José Luiz Maturana Pagnossim. 2018. Uma abordagem híbrida para sistemas de recomendação de notícias. Dissertação de Mestrado. Escola de Artes, Ciências e Humanidades, Universidade de São Paulo, São Paulo. DOI: 10.11606/D.100.2018.tde-07062018-101232

Alexandra Papoutsaki, Aaron Gokaslan, James Tompkin, Yuze He, and Jeff Huang. 2018. The eye of the typer: a benchmark and analysis of gaze behavior during typing. In Proceedings of the 2018 ACM Symposium on Eye Tracking Research & Applications (Warsaw, Poland) (ETRA ’18). Association for Computing Machinery, New York, NY, USA, Article 16, 9 pages. DOI: 10.1145/3204493.3204552

Alexandra Papoutsaki, Aaron Gokaslan, James Tompkin, Yuze He, and Jeff Huang. 2018. The eye of the typer: a benchmark and analysis of gaze behavior during typing. In Proceedings of the 2018 ACM Symposium on Eye Tracking Research & Applications (Warsaw, Poland) (ETRA ’18). Association for Computing Machinery, New York, NY, USA, Article 16, 9 pages. DOI: 10.1145/3204493.3204552

Alec Radford, JongWook Kim, Tao Xu, Greg Brockman, Christine Mcleavey, and Ilya Sutskever. 2023. Robust Speech Recognition via Large-Scale Weak Supervision. In Proceedings of the 40th International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 202), Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett (Eds.). PMLR, 28492–28518. [link]

Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language Models are Unsupervised Multitask Learners. Technical Report. OpenAI. [link]

Julian Salazar, Davis Liang, Toan Q. Nguyen, and Katrin Kirchhoff. 2020. Masked Language Model Scoring. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics, 2699–2712. DOI: 10.18653/v1/2020.acl-main.240

Junxiao Shen, Boyin Yang, John J Dudley, and Per Ola Kristensson. 2022. KWickChat: A Multi-Turn Dialogue System for AAC Using Context-Aware Sentence Generation by Bag-of-Keywords. In Proceedings of the 27th International Conference on Intelligent User Interfaces (Helsinki, Finland) (IUI ’22). Association for Computing Machinery, New York, NY, USA, 853–867. DOI: 10.1145/3490099.3511145

Richard Simpson, Heidi Horstmann Koester, and Edmund Lopresti. 2010. Research in computer access assessment and intervention. Physical Medicine and Rehabilitation Clinics of North America 21, 1 (2010), 15–32. DOI: 10.1016/j.pmr.2009.07.006

Kulvinder Singh, Sanjeev Dhawan, Nisha Bali, Ethan Choi, and Anthony Choi. 2025. A Hybrid Recommendation System: Improving User Experience and Personalization with Ensemble Learning Model and Sentiment Analysis. International Journal of Computing and Digital Systems 17, 1 (2025), 1–15. DOI: 10.12785/ijcds/1571024952

Stephanie Valencia, Richard Cave, Krystal Kallarackal, Katie Seaver, Michael Terry, and Shaun K. Kane. 2023. “The less I type, the better”: How AI Language Models can Enhance or Impede Communication for AAC Users. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (Hamburg, Germany) (CHI ’23). Association for Computing Machinery, New York, NY, USA, Article 830, 14 pages. DOI: 10.1145/3544548.3581560

Stephanie Valencia, Jessica Huynh, Emma Y Jiang, Yufei Wu, Teresa Wan, Zixuan Zheng, Henny Admoni, Jeffrey P Bigham, and Amy Pavel. 2024. COMPA: Using Conversation Context to Achieve Common Ground in AAC. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA) (CHI ’24). Association for Computing Machinery, New York, NY, USA, Article 915, 18 pages. DOI: 10.1145/3613904.3642762