Automatic Sequencing of Multimedia Content through Semantic Analysis
October 10th 2017 07:55 - 08:10
Proxem applied its artificial intelligence and semantic analysis technologies to audio and video transcriptions, as well as to user feedback on these programs. For this purpose, we worked on the optimization of the speech-to-text + semantic analysis pipeline: taken individually, these two tasks provide satisfactory results in most cases. What happens when we chain them on different families of contents? We have also implemented an efficient technique for splitting audio or video into sequences, with coherent topics, in order to achieve this operation automatically. Finally, we used this automatic classification for the recommendation engine. These improvements allow a better understanding of the desires of the listeners, in order to best meet their expectations and retain them.