In recent years, promotion of products and services using SNS has been popular, but a method for improving the popularity of specific contents has not been established.So far, many technologies have been studied to judge images and videos and recommend objectively correct tags (“mountain”, “lake”, “forest”, etc.).This is the world's first attempt to recommend a tag that enhances popularity (such as "reflection on the surface of a lake").This quantifies the strength that each hashtag affects popularity on SNS, and recommends any number of tags that are effective in improving popularity by referring to the hashtag given by the user. To do.

　This time, we let the system learn using about 6 images and the tags attached to them, and added the tags actually recommended by the system to about 2000 images and posted them on SNS. After 10 days, the number of views was about twice that of using only the tags attached by humans.Since the recommendation tags are originally based on the tags given by people, it has been verified by the subjective evaluation of about 2 people that the correct tags that match the contents of the images and videos are recommended.The time it takes to calculate the score is as short as a few seconds on a normal server, and you can re-recommend tags every day.

　Future developments include hashtag recommendation when effectively promoting products and services on SNS, improvement of click rate on EC sites, and support for creating impressive headlines in owned media and news distribution. It is said that it will be done.

Paper Information: FolkPopularityRank: Tag Recommendation for Enhancing Social Popularity using Text Tags in Content Sharing Services
* Scheduled to be announced at "The 26th International Joint Conference on Artificial Intelligence (IJCAI)"