2026-05-20T22:38:18Zhttps://keep.lib.asu.edu/oai/request

oai:keep.lib.asu.edu:node-1994992025-02-19T20:58:47Zoai_pmh:alloai_pmh:repo_items

199499 https://hdl.handle.net/2286/R.2.N.199499 http://rightsstatements.org/vocab/InC/1.0/ http://creativecommons.org/licenses/by-nc-sa/4.0 2024-12 47 pages Band, Devika Chavez Echeagaray, Maria Elena Suthapali, Vikash Gali, Mahesh Barrett, The Honors College Computer Science and Engineering Program This thesis presents the development of a user-centric AI-based writing assistant capable of analyzing, emulating, and modifying writing styles based on user preferences. The project focuses on enabling intuitive customization of generated text by allowing users to adjust parameters such as formality, humor, and tone through a dynamic user interface. The system integrates machine learning techniques to extract stylistic features from user writing samples, forming a comprehensive style profile. The study is designed to offer personalization and accessibility in AI-generated content. Users can interact with the model by providing their writing samples and adjusting sliders or selecting tone options (friendly, persuasive, or surprised). The backend processes these inputs to generate responses that align with the specified preferences while emulating the user's unique style. To validate the tool's effectiveness, an experience study was conducted, where the tool's output was evaluated for accuracy, user satisfaction, and adaptability. Results demonstrated that the model effectively adjusts key stylistic dimensions, with notable success in maintaining user writing patterns. However, challenges remain in capturing subtle nuances like rhetorical patterns and tone consistency under extreme customization levels. Insights from user feedback highlighted the importance of balancing quantitative evaluation with qualitative perspectives to enhance alignment and usability. The thesis identifies key lessons, including the significance of diverse training data, user-centered design, and the integration of advanced evaluation metrics. Future work will focus on refining the tool through style embeddings, reinforcement learning, multilingual support, and domain-specific adaptations. By combining AI’s capabilities with intuitive user controls, this research provides a framework for personalized writing assistance, empowering users to tailor AI-generated content to their specific needs and preferences. Transfer Learning Large Language Models (LLMs) Natural Language Processing (NLP) Generative AI Fine-tuning Prompt Engineering Custom Style Controls: Adaptive Text Generation with User-Specified Style Controls in AI Writing Assistants