| CPC G06F 16/45 (2019.01) [G06F 16/483 (2019.01); G06N 20/00 (2019.01); G06V 20/41 (2022.01); G06V 20/47 (2022.01)] | 20 Claims |
|
1. A method comprising:
receiving multimedia content comprising a plurality of frames;
obtaining a transcript of the multimedia content;
generating a plurality of keywords based on the transcript;
mapping the plurality of keywords across each of the plurality of frames, frame by frame, wherein the mapping comprises:
excluding one or more sentences from the transcript that do not contain a keyword from the plurality of keywords;
annotating the transcript with one or more timestamps based on the plurality of keywords; and
generating a keyword mapping comprising the transcript with the one or more timestamps and without the one or more sentences;
computing, for each of the plurality of frames, an importance score based on the keyword mapping and a chapter score, wherein the chapter score measures transitions between semantically coherent units of the multimedia content;
generating a ranking of the plurality of frames based on the importance scores and the chapter scores;
determining one or more top-ranked frames from the ranking that satisfy an importance threshold;
merging the one or more top-ranked frames into one or more moments; and
aggregating the one or more moments into a summarization of the multimedia content.
|