Abstract
Communication assistance tools may have difficulty determining an appropriate formality level for communications with new recipients where prior interaction history is unavailable. A technique is described for inferring formality by applying a large language model (LLM) to perform a lexicographical and semantic analysis of a recipient's email address string. A system can deconstruct the address into its local-part and domain-part, and the LLM can use its pre- trained knowledge to interpret socio, linguistic and organizational cues embedded within these components, such as role, based identifiers (e.g., "president") or specific domains (e.g., ".gov"). A resulting formality inference can be used to dynamically modulate the output of a text generation engine, which may facilitate the creation of more contextually appropriate message suggestions without a need for access to contact databases or manual user intervention.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Zimmermann, Nicolas and Zimmermann, Jonathan, "Formality Inference via LLM-Based Lexicographical Analysis of Recipient Email Addresses", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9986