Personalized content

Every day, millions of pieces of information are being updated on the Internet. The "rubbish" information also occupies a lot of space. They only interfere with the sight and waste time. In essence, the content itself can not say who is good or who is bad, in your opinion is rubbish information, sometimes for some people is not, the content is always only interested and not interested in both.

At this time, personalized recommendations based on big data seems to be a more effective method. With my powerful data collection, sorting, and extraction techniques, I've given my clients "keywords" based on past data and then recommended them to me for those keywords. The longer you read, the more the system understands me and the more accurate the recommended content.

To take the domestic personalized recommendation to do a rather good "today's headlines," for example, the first systematic collection of massive information, and then through data mining, intelligent analysis of the most popular every moment of the most worthy of users concerned about the information, and then Based on the Weibo analysis of the user's previous access to information, the establishment of a personal user model, a combination of the two, we can intelligently recommend personalized information for the user. But are these things really interesting to us? I make a big question mark.

One, unable to perceive interest, will only make us more and more narrow

Let me start by saying why I used the unofficial indignation after nearly a month of headlines today. Some time ago, the marriage incident on the article and Ma Yijie became the hot topic of Internet news. Today's headlines pushed me a number of push reminders, of course, these messages are indeed my interest, so most of the clicks. After that, all personalized recommendations to me are all kinds of entertainment gossip news. I use this software, the main purpose is to understand the hot events that occurred the same day, as for personalization is not very important to me, every day to go to the collection of major sites to see it again, basically has to meet my reading needs. But now full of screen entertainment gossip, for me a little information load. I like gossip, but not only in a gossip world.

Personalized recommendation seemingly understand me, can accurately guess what I like to read, not to mention whether this approach is scientific enough, the evaluation mechanism behind it is valid, even if it is scientific and effective, does this mean "I Just like to see the same kind of content "? If we always receive content that is in line with preferences, is likely to be in danger. This recommendation seems humane, but in fact castrated News. Users are living in a world of their own and can not perceive "new" things other than their interests. Without access to "new" things naturally can not cultivate new interests. Personalized news experience collected from your interest, but also determines your interest. Users eventually reduced to the bottom of the frog. In addition to making people feel the world news, but also inspire ideas. What worries me most is that personalized news experiences keep the mind from rolling out.

Second, the machine system, we can never know why we need to read

Today's headline every refresh will recommend 15 articles to me, of which about 2 content is my favorite, plus some of the most popular articles of the day, some degree of recommendation is not very accurate. The cost of a user refreshing a stream of information is very low, somewhat similar to Weibo, except for more and more content. But for specific users, a large number of not accurate recommendations caused by the experience is not good. This may not be relevant to their lack of basic data collection, but even if the future is accurate, this can not address the changing reading needs that result from changing scenarios. In fact, sometimes it is ridiculous that at some stage, we are concerned about this because everyone is discussing something. However, this does not mean that we are all interested in similar things.

It seems to me that this recommendation is related to the degree to which a person is deeply involved in a certain field. For an adolescent who has never seen a world in his life, all the AVs in this world are generally beautiful, and for a thousand readings, there is a story color Bright preference Petite well-proportioned contour soft Like being insulted but hateful death m Otaku, the recommendation is almost impossible to complete the task. This is not just a matter of finding content at such a level of detail, but also how you find out the need for such segmentation, as well as his different levels of arousal and tolerance of other non-optimal labels, The most crucial point is that we do not need to know his known needs but to explore his unknown preferences. This is the most unsolved part of the whole issue. The so-called unpredictable people are here. The machine can never be aware of why we like to read certain kinds of content at some point.

Third, more accurate recommendations, often bring the content of garbage

To say this judgment may in fact be controversial, but based on the personalized recommendation of big data mining, the first and foremost core is big data. How to have big data, which requires the system to collect huge amounts of data, and these huge amounts of data itself contains a lot of spam. To take the entertainment gossip I like to see this matter, I like to watch the gossip is only a few of the few stars, the rest of the stars are not my concern.

Today's headlines really can be precise analysis of my favorite gossip, recommended to me the news is indeed entertainment news, but these are not what I want. There are also over-recommended issues with popular articles, which leads to poor personalization. Step back and forth, even if the recommendations are all my topic of interest, the quality of the content is not very good. Sina Entertainment and an unknown entertainment station is to provide the same theme content, but for me their value is bad. I just want to see those high-quality content that I am interested in, but you have brought me so much noisy voice.

With the popularity of this kind of short message mechanism of Weibo, the text on the network is getting shorter and shorter, and people are getting impatient to read the long story. Of course, not everyone likes "short, flat and fast" information distribution. Our reading needs change all the time, and sometimes feel tired to see the short message, fast again; sometimes feel the need to charge, so want to see the depth of the article to understand the ins and outs of a matter, which requires reading the product Have enough flexibility.

The news does overload the information, but most people watch the news is the main way the hot news + their concern section of the news today + some of the more original type of news is not concerned about the original, so in fact the hot news and related sections Daily news is enough. In fact, every day we are helping themselves to make personalized screening, may today headlines to cover the content we usually can not cover, but our time has always been limited, you recommend a lot of news, we have time to see it?

Fourth, the lack of human factors, personalized simply can not talk about

Regardless of all the shortcomings we have mentioned earlier about today's headlines, our biggest problem is the lack of human factors. It seems to me that truly personalization can only be achieved by combining algorithmic-based content recommendations with recommendations based on social relationships. For example, I see a google glass science news, it is not cross-domain to Twitter who google engineers to participate in the project and its technology to share recommend to me

This is why I think microblogging irreplaceable. Today's top-level platforms are all kinds of news, it will never be able to capture the most subtle human emotion changes. A short 140-character Weibo may trigger our emotions and trigger some soul touching in our innermost feelings. We know that the Weibo accounts we are all concerned with are all living creatures. Because we agree with someone and like someone's thoughts, we achieve personalized self-improvement, which is absent from the headlines today.

Everyone is an era of media, and even a very small individual may become a news source. The former media were nodes of communication and now become specific individuals. The fact that today's headlines do not include these items is likely to create a situation where the flow of our information lags far behind others. We can not sense the changes in people's moods behind the news events, which is precisely why we want to read certain types of news. People are recommending and discussing each other's interests based on their commonalities, which we call "social." At present it seems that the existing "personalized reading" products, the problems you want to solve, such as Douban and Sina Weibo have been resolved well. The natural social recommendation mechanism of these products is more accurate and reasonable than the "push" of various products.

Fifth, interest is not dead, personalization will always need to face the plight

What the user does reflects his intention more truthfully than what he says, but for the purpose of speaking, the internet product adjusts itself according to the actual behavior of the user and self-adjusts with the traditional media based on the user's written comments Both are: To allow users to see more of what he is interested in, less to see what he is not interested in. But interest is not a static dead thing, the expansion of interest is to maintain the mental vitality of a person's premise. We say that the Internet is revolutionary, precisely because it reduces the cost of expanding interest and vision.

Personalization should be some sort of supplement beyond what we already know, but right now it just does not qualify for this complementary identity. Maybe one day we can do it, but now, whether it is based on the amount of data or algorithm, it only makes me feel that this is not a funny joke. I tentatively believe that what you are doing is a meaningful thing to do and I just do not believe it is something that is meaningful to me now. Compared with "personalization," I am more concerned about the same / similar information filtering. I think this is a more pressing and better solution.