EMNLP Findings2025
Adapting Large Language Models for Character-based Augmentative and Alternative Communication
Abstract
Users of Augmentative and Alternative Communication (AAC) may write letter-by-letter via an interface that uses a character language model. However, most state-of-the-art large pretrained language models predict subword tokens of variable length. We investigate how to practically use such models to make accurate and efficient character predictions. Our algorithm for producing character predictions from a subword large language model (LLM) provides more accurate predictions than using a classification layer, a byte-level LLM, or an n-gram model. Additionally, we investigate a domain adaptation procedure based on a large dataset of sentences we curated based on scoring how useful each sentence might be for spoken or written AAC communication. We find our procedure further improves model performance on simple, conversational text.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- Design and Analysis of Intelligent Text Entry Systems with Function Structure Models and Envelope Analysis cites A Design Engineering Approach for Quantitatively Exploring Context-Aware Sentence Retrieval for Nonspeaking Individuals with Motor Disabilities
Builds on8
- LoRA: Low-Rank Adaptation of Large Language ModelsEdward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu et al.ICLR 2022 · 18,833 citations
- DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding SharingPengcheng He, Jianfeng Gao, Weizhu ChenICLR 2023 · 394 citations
- "The less I type, the better": How AI Language Models can Enhance or Impede Communication for AAC UsersStephanie Valencia, Richard Cave, Krystal Kallarackal, Katie Seaver et al.CHI 2023 · 92 citations
- A Design Engineering Approach for Quantitatively Exploring Context-Aware Sentence Retrieval for Nonspeaking Individuals with Motor DisabilitiesPer Ola Kristensson, James Lilley, Rolf Black, Annalu WallerCHI 2020 · 76 citations
- Design and Analysis of Intelligent Text Entry Systems with Function Structure Models and Envelope AnalysisPer Ola Kristensson, Thomas MüllnersCHI 2021 · 75 citations
Related papers
- Learn Your Tokens: Word-Pooled Tokenization for Language ModelingAvijit Thawani, Saurabh Ghanekar, Xiaoyuan Zhu, Jay PujaraEMNLP Findings 2023 · 4 citations
- SkipWriter: LLM-Powered Abbreviated Writing on TabletsZheer Xu, Shanqing Cai, Mukund Varma T., Subhashini Venugopalan et al.UIST 2024 · 6 citations
- From Language Models over Tokens to Language Models over CharactersTim Vieira, Benjamin LeBrun, Mario Giulianelli, Juan Luis Gastaldi et al.ICML 2025
- Hierarchical Autoregressive Transformers: Combining Byte- and Word-Level Processing for Robust, Adaptable Language ModelsPit Neitemeier, Björn Deiseroth, Constantin Eichenberg, Lukas BallesICLR 2025
- Giving Meaning to Movements: Challenges and Opportunities in Expanding Communication by Pairing Unaided AAC with Speech Generated MessagesImran Kabir, Sharon Ann Redmon, Lynn R. Elko, Kevin Williams et al.CHI 2026 · 1 citation
