Vernacular recommendation system

If you want to go out to eat hot pot, you can go around all the hot pot restaurants outside your house, choose the right one according to the price and taste, eat, pay and leave. If you live in a food city, you can follow the signs at the door to find the floor and then follow the directions to the hotpot restaurant where you can order a meal. If you don’t want to go out, you can open the food APP on your phone, type “hot pot” in the search box, place an order, pay and wait for the delivery. The above several cases are in the case of users have clear needs, through the search engine to find what they need. So what if the user doesn’t have a clear need? Let’s say it’s a Sunday afternoon and you want to watch a movie, but when you open YouTube, you’re faced with a variety of movies and you don’t know how to choose. At this point, you have two options, one is to consult movie-loving friends around, or use an automated tool – recommendation system.

The application of recommendation system in social media

With the development of information technology and Internet, human society has stepped into the era of information overload from the era of information shortage. This is a challenge for both producers and receivers of information: producers want to get the most exposure for their content, and receivers want to receive the most valuable information. To solve the problem of information overload, numerous scientists have proposed solutions, most representative of which are classified directories and search engines. Yahoo, a famous Internet company abroad, started by relying on classified directories, sorting famous websites into different categories, so as to facilitate users to find websites according to categories. However, with the continuous expansion of the Internet scale, classified directory websites can only cover a small number of popular websites, some unpopular and minority websites can not be covered, which is increasingly unable to meet the needs of users. Search engines were born. The search engine represented by Google allows users to find the information they need through search keywords. The representative company in China is Baidu.

Google Search interface

Search engines need users to provide more accurate keywords, so when users cannot accurately describe their search needs, search engines are powerless. The birth of recommendation system solves this problem well. Different from search engines, recommendation systems do not require users to provide clear information needs, but actively recommend information that users may be interested in by analyzing their historical behaviors and modeling their interests. From the user’s point of view, search engine and recommendation system are two different means for users to obtain information. Search engines meet the active search needs of users when they have a clear purpose, while recommendation systems can help users find new content they are interested in when they have no clear purpose. Whether on the Internet or in real life, these two ways coexist in large numbers and are complementary. Take shopping in the mall as an example. If we know clearly what we want to buy, what model, style and color, we can just go to the store and buy it directly. This is called “search”. If we are not quite clear about the size, model, style, price and so on of what we need, we need a shopping guide to tell us which goods meet our needs, which is “recommendation”. From the point of view of objects, search engines play the Matthew effect, recommendation system is the long tail effect. The Matthew Effect is a phenomenon in which strong people get stronger and weak people get weaker. This theory comes from a parable in the New Testament Gospel of Matthew: “To everyone who has, give twice as much, and he will have more. If you don’t have one, take away even what you have. “This is similar to the 80-20 rule. Take Baidu’s search click as an example, the higher the ranking of the search results, the more likely users are to click, the lower the ranking results and the results after turning the page are less likely to be clicked. That’s why Baidu’s ads make so much money, and why companies are scrambling to do SEO. The higher the ranking, the more likely it is to be clicked, “the stronger the stronger”. The Matthew effect is the long tail. Chris Anderson, editor in chief of Wired magazine, wrote “The Long Tail” in 2004 and “The Long Tail” in 2006. The long tail is used to describe the distribution of hot and cold items in economics. In the Internet era, because network technology can make people access to more information and choices at a low cost, more and more “forgotten” items are paid attention to again on many websites. In fact, everyone’s tastes and preferences are not exactly the same as those of the mainstream population, Chris points out: The more we discover, the more we realize that we need more choices. Mainstream products tend to represent the needs of the vast majority of users, while long tail products tend to represent the personalized needs of a small number of users. However, the activation and utilization of the long tail resources is exactly what the recommendation system is good at. Users are usually unfamiliar with the long tail content and cannot actively search for it. Only through recommendation can these unfamiliar content attract users’ attention and arouse their interest. For producers, revitalizing the long tail of resources is also critical. Just think about it, if an enterprise only relies on a certain type of commodity to attract popularity, then when this kind of commodity is no longer popular and the new commodity has not been supplemented, the enterprise’s earnings will be subject to great fluctuations. Only by relying on the recommendation system to fully explore the long tail, meet the personalized and differentiated needs of users, expose the long tail content at the appropriate time and attract new popularity, can the healthy and stable operation of enterprises be maintained. So how does a recommendation system work? Take our movie watching as an example. When deciding what movie to watch, we will probably experience several ways:

1. Ask your friends. We can ask our friends who love watching movies about the best movies they’ve seen recently, or we can wait on moments for recommendations. In recommendation systems, we call this social recommendation, where you recommend things to yourself through your friends.

2. Start with your favorite actor or director and search for his or her other movies you haven’t seen yet. This approach is to look for films that are similar in content to those you have seen before. This approach is called content-based recommendation in recommendation systems.

3. We can also choose our favorite movies based on douban ratings and netizens’ ratings. This approach is called collaborative filtering based recommendation.

The above is a brief introduction to the recommendation system. In the final analysis, recommendation system is one of the ways to connect users and items. It can help users find what they are interested in in the environment of information overload, and can also recommend information to users who are really interested, so as to give full play to the maximum value of information and meet the choices of different user groups. Reference: Xiang Liang. Practice of Recommendation System [M]. Beijing: Posts and Telecommunications Press, 2012.

Related reading:

Want to learn about recommendation systems? Look here! (2) — Neural network method

Want to learn about recommendation systems? Look here! (1) — Collaborative filtering and singular value decomposition

How to realize automatic online, operation and maintenance of intelligent recommendation system?

Getting started with recommendation systems, a list of knowledge you shouldn’t miss

For more information, please search and follow the recommendation wechat public account (ID: DSFSXJ).