Abstract
Computing devices provide multiple applications that may incorporate similar functionality. Recommending a particular application for a current user task is necessitated when a user relies on the operating system or a virtual assistant to recommend the application to use. Current feature recommendation techniques rely on predefined descriptions and criteria, which can be rigid and inaccurate. Such techniques also do not take into account user preferences regarding interaction modality for the tool to use. This disclosure describes techniques that leverage a large language model (LLM) to provide feature recommendations (app recommendations) accurately and that take into account user preferences regarding modality of interaction. Per the techniques, an LLM is used to generate dynamic service descriptions, establish quality criteria, and continually update both the service descriptions and the quality criteria based on user interactions and feedback. The techniques enable accurate and adaptable feature recommendations, enhancing user experience and promoting the use of the most suitable tools for any given task.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Ansari, Saad; Brooks, Mindy; Burrows, Rachel; Cao, Ann; Carless, Jessica; Graves, Rod; Greenwood, Ben; Gundersen, James; Huang, Allen; Jönsson, Mårten; Jung, Hyuniee; Li, Luke; Lu, Nikki; Mcgeachy, Heather; Neiswander, Laura; Park, Angie; Salois, Rebecca; Squire, Patrick; and Tsurkan, Artur, "Modality and Quality Aware Feature Recommendation Using a Large Language Model", Technical Disclosure Commons, (June 10, 2025)
https://www.tdcommons.org/dpubs_series/8222