A Free Trial That Lets You Build Big!
Start building with 50+ products and up to 12 months usage for Elastic Compute Service
Intermediary transaction SEO diagnosis Taobao guest Cloud host technology Hall
The first edition, which was launched at the end of February, is now nearly 10 months old. For a long time did not update the blog, today to launch a "No network" of this product opportunities, and we explain that there is no way to go.
The idea is to make the Internet smarter and more personalized, and to achieve it is based on "1 + 1 > 2″."
1 + 1 Why is greater than 2?
The author of the first contact with the study of the recommendation system is from the instructor introduced an article called "Collaborative filtering ...", Chinese translation into collaborative filtering. At that time the first reaction was surprised, the recommendation is to give you more information you are interested in, how to become a filter? It was later understood that the information you are interested in is filtering out information that you are not interested in, but not the same angle. But really stay in the author's brain is the meaning of collaborative filtering, a lot of things exist alone when not much use, combined to produce new value, such a simple truth can also be used in the algorithm. The principle of collaborative filtering is very simple, it is based on the similarity of people, things have similar, to recommend books to you, only to find people with similar tastes, the book they like to read and you have not seen the recommendation to you, very likely you will like (because you taste similar). Two interested readers, each one of them interested in the article, the two clicks alone may have been no use, but combined to be able to recommend to each other. A single neuron (neuron) does not work, but countless neurons are connected in a meaningful way to become our brains. There are many similar examples in daily life. Because of synergy, 1+1>2 will be greater than numpad this seemingly simple logic into the soul: from the earliest company name "Two Wood", to the recommended algorithm of collaborative filtering, and then to today to launch the web-based interconnection of the web.
From the relevant article plugin
Did not find a few months ago developed a related article Plug-ins, by the webmaster's favorite, many well-known domestic bloggers are in use. Related article plug-ins have been in existence for many years, casually in WordPress search on a few hundreds of, do not look for why to do it again? The most fundamental reason is that plug-ins have their limitations, innovative space is very small (here Plug-ins refer to the general attached to the blog frame such as WordPress, Zblog, etc. plugins).
Common Traditional practices
Related article plug-ins the most common approach is to rely on the article's label (tag), the same label means that the two articles to a certain degree of relevance, the more the same label on the more relevant. This is easy to understand, the algorithm is also very simple, generally to the Web server does not cause too much pressure. Based on similar ideas, some plug-ins also consider the article category, content, time, and so on, but these all have a same fatal point.
The drawbacks of traditional practices
General plug-in algorithm processing is completed in the plug-in end, which means that the plug-in runs very fast, otherwise it will affect the speed of the Web page loading, which is also the majority of webmasters do not like to install too many plug-ins the main reason, lest affect speed. In this way, the innovation of the relevant articles will be greatly limited. This points out some of the relevant articles can be improved, but because the plug-in environment is restricted and difficult to achieve.
1. Introducing behavioural data will help readers find more good articles. Behavioral data refers to what you often see on Taobao or Amazon, "buy this and buy that, or see this also look at that" function. In the recommendation system, the most classical collaborative filtering method is difficult to realize because of the long processing time.
2. The label weight of the article may not be the same, an article has more than one label, but may be inside a tag to really represent the theme of this article, such as the ability to identify the weight of different tags will help find more relevant articles.
3. Processing pictures, video and other heavy work usually takes a long time, it is difficult to complete directly on the plug-in side.
4. Perhaps later related articles are no longer limited to related articles, but recommended, users are interested in but may not be related to what is being seen. Perhaps the relevant article is no longer limited to the station, but can be any article on the Internet.
Creative approach: Cloud related articles Plugin
Some might argue that the value of this improvement is limited and may not be worth doing. But when Google started to do search, the search industry was not bullish. In order to be free from the limitations of the general Plug-ins, the model of cloud computing has not been adopted:
The main task of the plug-in side is to send a request to the diskless server, and then return the relevant article. This approach will not cause any pressure on the site owner's server, and then give a lot of space to innovate. No search can be done to any complex algorithm to calculate the correlation, to do the time needed for the image of video processing, just to cache the results of the line. There are several foreign countries to adopt a similar pattern, but the relevance of the Chinese Web site article is always strong poor. Because there is room for innovation, there is no relevant article plug-ins now to the site to take an average of 10%-20% page traffic.
Prior to the development of the relevant article Plug-ins is to "find the network", this is a long time we prepared products, but also a bold experiment, I hope that all the webmaster can participate in this experiment.
What is no network?
No search network is composed of a Web site, in this network, the main site can request a connection to any Web site, once the other party through, linked to the site will share resources to create more value. At this stage, the linked sites will display their respective contents in the relevant articles. No-looking will ensure the relevance of the article, the exchange of the flow of fairness, cheat and so on a series of technical issues.
What is the value of no network?
Reader's perspective: reading is no longer limited to the same website. When readers read the article "Google intends to spend 6 billion on Groupon" on a website, he may go on to see "Groupon has rejected Google's takeover," but this article is on another site. Before the website is connected to the website, the reader will be limited to the information of a single website, and the information can be complementary.
Stationmaster's angle: The reader quantity soars rapidly. Suppose your site has 10,000 readers, connect to another site, and the total number of readers becomes 20,000, with 100 connections, and the total number of readers becomes 1 million. Even if there are overlapping readers, the scale of the network effect can be surprisingly large.
Marketing point of view: When the small webmaster know the site just started when the promotion of how difficult, do the most is to exchange links with other sites, visits to the site, send IP, everywhere to leave their own site links and so on. Recently saw a network marketing forum unexpectedly can in a two week jumped became the domestic first 500 website, visible stationmaster to this demand has how high. The site content of good or bad and webmaster understand marketing has no direct relationship, to the webmaster to spend so much energy to do to a certain extent is a waste of resources, in the past because there is no such as no network such a platform, if you can let the webmaster to save these times directly in the creation of better content, I believe the entire internet will become better.
Several websites are invited to participate in this experiment, cut a picture to see:
Why do we have to do no network?
It is generally believed that only large companies will choose to do the platform. There is little resources, the team is very small, but we believe that synergy, believe 1 + 1 > 2, I believe that the webmaster and webmaster can form a very powerful force.
When we are ready to launch the network, it happened that "Wired" magazine founder Kelly came to visit the country, like his passage:
Kelly: I still use the swarm as a metaphor, the behavior of a colony is not found in the individual bee. But the overall behavioral characteristics of the colony come from many individuals, such as the memory capacity of the colony, which, if we use a certain lever to measure the memory capacity of the colony, is longer than the memory of the individual bee.
The future of the "one" system is similar, that is, all 3G or mobile phones, even when a large network will emerge completely different behavior characteristics. These things are undetectable by a single mobile device.
The individual is unconscious, and the group finds its direction in the overall runaway. This is the change that the web brings to the world.
Update: No network has now accepted the application, address: http://wumii.com/site/index.htm
Start building with 50+ products and up to 12 months usage for Elastic Compute Service