How is the content accurate to recommend to the user?

  If we refer to the product/video collectively as the content of a website.

The way the

content is pushed, searched, sorted, and filtered is categorized into the following four forms, more or less, and coexist.  . The user chooses this one way the user has the biggest initiative, the website through the design clear classification navigation and the station search system, lets the user through own intention, searches the website to find the content information which he needs.   Advantages: Clear, simple and effective. Disadvantages: Whether accurate information can be found, on the one hand, depending on the user's classification of information and search keywords of the organization, on the one hand, depending on the site of information classification and keywords organization. If the two parties to the same content of the needs and understanding of inconsistent, an information content, the site is classified as a category, users think B category can be found, then must be reactive and return. So this information should hit a, b two tag, grouped into two categories. The same is true for search, where information that may be queried by the user is fully available in the search index. Finally, a lot of classification and miscellaneous search information accumulated, but also from the precision into a blur. So people found that the major search engines on the word more and more long.   Summary: As the most basic way, no site will not provide this function, at least have a classification of information navigation.   Two. Audit matching audit refers to the content of the audit screening, such as through the scoring system, browse volume, click Rate, time and other factors, and a large number of similar information for sorting and weight distinction, to determine the content of the user needs level, in the Category list or search results, Display the most likely user needs information first or most clearly to the user.   Advantages: For the first Way to supplement, through the rules of information sorting and filtering, reduce the search cost of most users. Disadvantage: This is a minority of the majority of the matching method, that is, the default needs of information users are "most users", ignoring and ignoring the "minority user" needs. The scoring system also has systemic problems, so that users can score, and will not produce idealized scores, professional ratings and lack of democracy. As for the issue of commodity scores, I have discussed in another article--imperfect scoring system of existing commodities in addition, YouTube has concluded that the distribution of ratings is not idealized.   Summary: In this matching system, maximum demand is met, the minimum demand is ignored, from the overall profit and loss is the best solution, but when the personalized needs of this society more and more strong, the long tail of the user needs are not obvious order, is a small number of users have a small number of different needs, And they add up to be a long tail that can not be neglected, how to push the content information to them.   Three. Personalized matching through multi-dimensional understanding of user preferences, the user's information needs to match. People who like to drink juice through the word "apple" first look at the "apple Juice" related information, and a "fruit powder" through the word "apple" first found "Apple Computer." How to define a userSexual dimension, there are generally two ways. One is called "Birds of a Feather": it is estimated by what it likes. Watercress is very good at personalized matching, it can give users to push very accurate books, movies and so on. Therefore, watercress needs to understand what type of content you like, you need to constantly accumulate your favorite content, it will be more accurate to similar content to you.   Another is "flock together": By your social circle to predict. The traditional way of information transfer is through the "group" of people to carry out, information sharing is the first in the Circle and circle, the interests of the same people spread, so this is a simple method, through the Internet technology to improve the efficiency of communication. Watercress will also inform users of his friends in the attention of things and the collection of information, the most typical or similar to the SNS network such as the basis of social networking. Your friends obviously know you better than computer algorithms, and they actively recommend it to you, or their own favorite information, and believe you will be interested.   Advantages: More accurate matching, but also more effective shortcomings: Matthew effect (Matthew multiplying), the better, the worse the bad, the more the more, the less. For example, I subscribe to Google Reader in a large number of ued-related blog, and it automatically recommended ued related blog to me, but I have subscribed enough, I would like to see some write gossip text feeds, but it is difficult to find. I also can't find the right blog by the keyword "write gossip little Feeds". Summary: Most accurate, but easiest to walk slant.   Four. No difference pushes the so-called no difference is, whoever you are, will give you the same message.   Advantages: Easy, direct, simple disadvantages: Easily be user aversion summary: also has the effect   in general, I summed up in these four ways, each has advantages and disadvantages, as to how to use, in different scenarios there are different specific ways, this more, that more.   Because of the previous article "Tudou's Hunger and creativity," signed for Cronish's pertinent comment that "potato-related video recommendation mechanism is problematic, do not solve the core recommendation mechanism, but only to make a page location changes are failed product changes." "So I would like to take this reason to study how to best push content to match to the user."