Techniques are presented herein that support a "HiMEG" system, a framework that helps create meeting notes of multiple granularities for meeting invitees so that they can refresh their memory or catch up on any meeting. Such a system comprises a Segmentation Engine that may divide a meeting transcript into separate sections representing the different topics that were covered during a meeting. Such a system also comprises an Attention Correlation Analyzing Model that may be used to capture the attention correlation between different meeting notes that were generated from the discovered topics, which is useful in a Meeting Note Summarization Model that may assess which meeting notes are most similar. Under such a system, one effective summary may be formed based on the most similar meeting notes and the process may be repeated until there is one overall summary of a meeting. In the end, a user may read the high-level summary of a meeting and then dive further into the specific contents of the general meeting note based on their interests and needs. While the above-described framework was originally developed for generating meeting notes, it may also be applied to any text input such as speeches, action scripts, and training scripts.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.