The paper presents ‘Rambler’, a novel LLM-powered interface that improves text dictation on mobile devices. Rambler offers two main functions: gist extraction and macro revision. Gist extraction generates keywords and summaries from dictated text, aiding in review and interaction. Macro revisions allow users to rephrase, split, merge, and alter dictated text without pinpointing specific editing locations. In a study with 12 participants, Rambler outperformed current speech-to-text editors, offering enhanced user control and supporting a diverse range of user strategies.
Publication date: 2024-01-19
Project Page: https://doi.org/10.1145/nnnnnnn.nnnnnn
Paper: https://arxiv.org/pdf/2401.10838