Introduction: How does Pumpkin Movie solve the pain points of long and short video business through personalized recommendation?
company background
Beijing Xiaoming Zhumeng Data Service Co., Ltd. is a new media company specializing in the operation of online self-produced dramas and movie copyrights. Its APP-Pumpkin Movies, is a leading vertical video product focusing on high-quality film and television operations in China. It is aimed at the majority of middle-class elites and provides high-quality products that are different from theaters and other video platforms on mobile Internet, OTT and other clients. The customized and differentiated content operation service creates a high-quality professional audio-visual feast for the majority of users.
Business pain points
Pumpkin Film, as a video service platform focusing on high-quality film and television dramas, integrates long and short video services, and has the following typical industry pain points:
1. Long video recommendation needs to take into account multiple goals: different from other video industry scenes, movies and serials are typical long video scenes, with concepts such as duration and number of episodes. The index to measure the recommendation effect is not only the click rate, but also the user. The length of stay, the number of plays, etc.
2. There are high requirements for the accuracy of recommendation: movie recommendation, in addition to the most common collaborative filtering algorithm based on behavior calculation in the recommendation system, also needs to combine the material, theme, core content, and fragment content of the movie to include semantic understanding , In-depth interpretation and other features to enrich the material portrait and improve the accuracy of recommendation.
3. Exposure filter settings for long videos : The browsing of long videos needs to be carried out with a lot of leisure time. Sometimes users are not not interested in the recommended content, but lack a complete time period to complete a viewing. In this case, it is necessary to consider the user's repeated access requirements and set a reasonable exposure filter time.
4. Diversified combination of long and short videos : Compared with long videos, the short video section focuses on the wonderful pictures in the movie, bringing users a unique viewing experience. It is recommended to integrate different video types according to the needs and habits of users. Features, goals, multi-faceted arouse user interest, enhance user stickiness, and increase user stay time.
solution
1. On the one hand, it integrates film theme introduction, film reviews and other content, and automatically marks through semantic analysis to enrich film tags. On the other hand, through the professional operation and editing team of Pumpkin Film, the description of the movie is visualized through label descriptions. These labels are input into the recommendation system as important features of the material, which improves the accuracy of the recommendation and greatly improves the recommendation effect.
2. Taking into account the factors of user browsing time, through algorithm test analysis, it is calculated that when the exposure filtering time is 3 days, for the user group, the content can be appropriately repeated to reach the effect of improving the recommendation effect.
3. Pumpkin Movies has created a special area for users to recommend movie clips (short videos), which can not only provide users with a unique viewing experience for their fragmented time, but also quickly present movie highlights and guide a complete movie viewing behavior.
4. Through the intelligent algorithm access of long and short videos, the user's interest is fully opened up, and according to the user's habits, the theme of the movie they like to watch, to achieve further accurate recommendation. At the same time, taking into account the diversity of recommended movies, let users have new discoveries, and promote the increase of viewing time.
product value
Alibaba Cloud Smart Recommendation is the Pumpkin Movie APP, which provides recommendation services for the "Pumpkin recommended for you", "Recommended according to the "XX" you have seen, and "Similar Movies and TV" sections, which can be personalized according to the habits and preferences of different users recommend.
Mainly through the content industry to undertake long and short video recommendations, integrating multi-objective optimization, including stay time, click-through rate, consumption indicators, etc. In the Guess You Like section, combined with the long-term and short-term interests of users, intelligently generate user portraits and provide customized recommendation experience; in the related recommendation section, combine movie semantic analysis, content association and user behavior analysis, calculate the most relevant movies, and make quick recommendations.
This article is the original content of Alibaba Cloud and may not be reproduced without permission.