As a senior user of NetEase Cloud, I found a lot of unpopular but really good treasure songs on NetEase Cloud. Treasure niche song recommendation counts as a differentiated advantage of NetEase Cloud Music, which is rarely seen on other whatsapp Database music platforms. Today, I will share with you the story behind the unpopular treasure song, and see how the NetEase data product team realizes batch mining and cold start of niche high-quality music by defining music connoisseurs . First let’s talk about why we don’t hear niche songs in most cases. From the perspective whatsapp Database of the recommendation algorithm, because the user behavior data of long-tail content is relatively sparse, the algorithm has a high probability of misjudging.
And recommending some content with low popularity and unpleasant sound after misjudgment is very harmful to the user experience. Therefore, in general, the algorithm will whatsapp Database prefer to recommend popular content, because the amount of data is sufficient and the confidence is high. This will make the entire platform pursue deterministic low-to-medium returns , but this will lead to a deformed ecology and convergence of music whatsapp Database tastes. Long-tail content: It can be understood as content that is niche, unpopular, and has a low penetration rate to the crowd. Long tail refers to the position of the tail in the probability distribution graph. NetEase Cloud Music.
How to discover music connoisseurs through data, and then find treasured niche music The algorithm that focuses on recommending mature songs is very unfriendly to new creators. Such a mechanism will make new works and niche works in a whatsapp Database subdivided circle not favored by traffic, without the opportunity to show their faces, and naturally unable to stand out. From the perspective of the platform side, if the new content cannot be cold started well, the traffic will always focus on the songs with high popularity, which is very unfavorable for the whatsapp Database long-term healthy content ecology. For example, Kuaishou has suffered from excessive concentration of traffic in several live broadcast families.