Abstract:

Abstract: The main source of revenue for the YouTube platform is advertising. YouTube acts as an inter- mediary between advertisers and viewers, which means that the primary goal of the service is to max- imize the number of displayed ads by increasing the average daily time spent on the platform by users, but also to precisely target ads to the right audience groups and create a favorable environment for the presentation of the aforementioned ads. The above-mentioned goals can be effectively realized on such a large platform only with the help of an efficient recommendation algorithm, which necessitates the need for its constant improvement. The article proposes the use of comment analysis via natural lan- guage processing to select the right content to promote in order to generate higher ad revenues due to their greater effectiveness and thus value. Evaluating the sentiment of comments is proving to be a ma- jor challenge due to the difficulty in recognizing context, sarcasm, irony, and mockery. Associating polar- ity with kurtosis and skewness significantly facilitates the recognition of negative and positive com- ments.