Abstract
Large language models (LLMs) have shown substantial improvements on text summarization and classification benchmarks over prior techniques. LLMs can be used for topic detection for online video which can help improve video recommendations. This disclosure describes techniques to parse comments associated with a particular video in addition to video topic detection. The automated parsing enables acquisition of auxiliary information about the context in which the video is made available. The topics/keywords can be complementary to or can extend upon topics detected from the video and/or video metadata. A combination of the video topics and comment topics can be utilized for video recommendations and can identify potential audiences that may not otherwise be matched to a video.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Dimarco, Mason Henry, "LLM-based Comment Summarization and Topic Matching for Videos", Technical Disclosure Commons, (May 16, 2023)
https://www.tdcommons.org/dpubs_series/5900