Abstract
Teams often struggle to evaluate the quality of features that process the content of virtual meetings, e.g. meeting summarizers. For such evaluation, one needs a large, realistic dataset of meetings but this is often difficult to obtain as real meeting contents are sensitive, privacy-restricted data. Proposed herein is a system that generates synthetic yet realistic meetings that mirror how people talk, with varied personalities, moods, and interactions. The generated meetings can follow the specified overall distributions of key parameters like duration or the number of attendees (thus leading to a broad, representative dataset for general evaluation), or they can replicate a specific meeting case if fixed values are specified for such parameters (thus enabling more targeted evaluation).
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Karetka, Gregor; Ondrejova, Viktoria; Sucik, Samo; Gerolemou, Zoe; and Skala, Daniel, "SYNTHETIC MEETING GENERATION SYSTEM WITH OPTIONAL REFERENCE-GROUNDING", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9892