What is recommendation?

The understanding of recommendations is always profound, and it is easy to go into similar misunderstandings. Next we will take movie recommendations as an example to discuss my personal understanding.

1. The recommendation is not a rating.

Rating movies is generally required for recommendation, but "Recommendation! = Rating ". Otherwise, you only need to recommend the most rated movie to the user. However, for those who love terrorism movies, he does not like other types of movies that you recommend to him, even if the rating of other terrorism movies is lower, you will also want to see it (even if you have finished reading it, you will vomit again ).

2. The recommendation is not to predict the personal score.

For those who like terrorism movies, he will give a high score to "Dolphin Bay" (documentary) and give a low score to "XX Ghost King". However, A high score does not mean you want to watch the movie. If you also recommend the two movies, you may choose to watch the latter. -- What I want to see is not necessarily something that I think highly. (For example, I will praise people with high moral character and save their lives, but I do not necessarily want to do this myself.High comments and desires are two different things.)

3. The recommendation is not necessarily just a recommendation of similar things.

When I listened to music on Douban TV station, I marked two martial arts songs I liked. I immediately recommended ten similar songs. It will make others get bored. The same is true for those with high ratings. I like it. It may be a long-term interest, but it does not necessarily mean that I have to listen to ten of them now.You cannot eat the same things at once, so that different types of data can be kept fresh for a long time.

4. The freshness in the recommendation is very important.

Freshness can be divided into two types. FirstRepeatable freshnessFor example, when listening to Douban radio station, you only need to mix and match the music and change constantly, even listening to the songs I have heard (or even like this). SecondBasically do not repeat freshFor example, watching a movie or reading a book. Few people read a movie again. At this time, we will find new movies that we haven't seen.

5. Recommendations should be valuable to users.

This is mainly intended for reading books. Especially in Professional Books, when readers want to know more professional fields of knowledge, they will be increasingly involved in niche fields. How can we find out what kind of books will people who are reading this book continue to read "?

6. Recommendation should be of commercial value.

Hot recommendation and long tail recommendation are both needed. Without considering social responsibility, we aim to create the greatest commercial value.

References:

Http://www.52ml.net/16040.html