Follow people, or follow content

The world trend, together for a long time must be divided, long time will be combined. Since the Web site, digital information has started to rise, until we have a search engine, this information is not well organized. Ever since the concept of feed flow, information has suddenly ballooned, and until we find the right tool, the information cannot be pushed to the right person. All along, people never stop to integrate all kinds of flow information like capillaries. In particular, some excellent websites that are already far away from the wall, have made a lot of good solutions, such as Facebook http://www.aliyun.com/zixun/aggregation/9105.html "> SNS website attempts to organize the feed stream through the relationship of the real friend, the all-encompassing FriendFeed attempts to put all the feed information to one place (domestic similar has this year Zhang writes the Kanrss), the two years became the upstart Twitter to make the flow information generation easier, Receive more convenient, follow is the income.

The above information solution is through follow people to achieve, and follow content is undoubtedly another feasible way. On the content of the classification organization has been for many years of research and practice, in the Zheng of the topic Engine Blog in a very detailed review, and the follow of these categories to get a category of information updates, as I know, still not very rare. More well-known such as Google News and Google News, the former belongs to the content of the classification of the organization, the latter belongs to the content of the classification of the subscription or say follow. Relying on Google's strong search capabilities, the novelty and breadth of information is assured, but because the product positioning is not to make a detailed classification of content, so the classification is more coarse, just some such as the portal site of the rough categories.

Cutt.com, which has recently been online, wants to push this classified information to the extreme, a company that boasts semantic web technology as its underlying analytics engine. It's on the line, first thanks to the country, or maybe we can see it a few months earlier. Because of Ku Wendong's introduction, I was able to have a certain experience in the early days of an organization with a big ambition but not very mature information. This is a very thoughtful newborn, but now some places are not very mature.

Products and Interactions:
An engineer-led company is easy to make a astounding creative product, but it is also easy to expose some product design and interaction problems, perhaps because intelligent engineers are always difficult to put themselves in the majority of uninformed users of the situation.

Cutt very good point is that the user uses 0 cost, any one user opens to use, does not need to register, also does not need to carry on any information to fill. I can do most of the operation in the anonymous state, there is no sign up without the situation of the collection of articles incredibly can save, very exciting! But the terrible thing came, and once I logged in, I found that all the data I had thought had been recorded was gone. Well, I might have thought it would automatically import anonymous information into my account. But then I thought, if it did, I'd be more scared. In fact, I mean: I didn't get any hint of the case, I don't know where I just was in the data. For ordinary users, there may be another problem is to change the browser, the data is gone, users do not feel that they are wrong, they will only think that you have lost their data. The same situation is that I wrote the article feedback, the same without a trace, although I know Cutt's database must have records, but most users will also think you lost TA data. Although I do not do products, but I think there is a principle: users contribute to the data TA will be able to find, otherwise the consequences are not only TA no longer willing to contribute so simple.

The other is the content, because it is a newborn, the content channel is not perfect enough, such as photographic equipment content is not, because I recently focused on this, so I saw this, but the other side of the estimated channel is still missing. For example, I found that Cutt do not like entertainment, because a lot of entertainment suites are still empty:). To play the content of the site is one of the core competitiveness is the update speed of information, and Cutt information update speed is still to be improved, I saw at 10 o'clock in the latest article or five o'clock in the afternoon more, do not know is subject to the crawler or algorithm processing speed. In addition, the presentation may be improved, simplicity is a way, but if only the news as the main carrier, add some visual elements may be more attractive.

Technical aspects:
Preview: I like Cutt's article preview so I don't have to wait for the entire page to load. But I do not know whether it can be further filtered, using Text Digest technology, the main content in a few words to pass out. For today's fast-food social consumption habits, this is undoubtedly a very competitive feature. I've even considered the abstraction of the abstract information by people, which is part of the swarm intelligence.

Source: According to my observation, the current Cutt still take the Web page this kind of unstructured information primarily, the source also mainly is some large and medium-sized portal or the information website. In fact, in the current era of the mashup, RSS feeds on the network a lot, if you can make full use of blogs and some web2.0 web site output RSS semi-structured information, the source of information will certainly be richer, more analytical certainly stronger. Of course, I reckon Cutt must have this content, but there is no more release.

Google Reader: Once I was a Google reader of the heavy users, almost every day, but also subscribe to a large number of interested or interested in RSS feeds, and the evil thing to classify them as algorithms, networks, science, it information, amateur and so on channel. But then I had to look at it more and more, allowing the 1000+ to stay in the upper-left corner forever. The reason is because the content of follow is dead, and follow is alive, there are emotional factors in the inside, so if a person does not have more time, TA will be more inclined to brush their own meager, rather than cold content. But the need for information is still there, so I am now more in consumption through friends filtered information. If an algorithm can have better filtering ability, I am still willing to use, especially personalized information recommendation. Because the neighbor recommendation is for everyone, not your own, and in this respect, the machine can do better.

Personalized Information recommendation: Although Cutt has not yet, but I know that there will be, now is only to spend a cold start of the user information, the more contributions, the more harvest. But information personalization is a much more difficult task than information organization, besides considering the semantics and relationship of content, it is now more complicated than the content to solve this problem.

Thought level:
Finally to a point of virtual.
The use of collective intelligence: not only the use of user implicit feedback data to the socialization of the recommendation is so simple, it is more important to include the user's explicit and voluntary contributions to the content. The objective authority of Wikipedia, for example, comes from the numerous spontaneous editorial actions of netizens, such as the construction of the benchmark freebase of the semantic web, which relies on the contributions of a large number of volunteers. Completely rely on the enthusiasm of the user obviously not, especially in the domestic Internet environment people are often more willing to take rather than contribute, how to make users happy to contribute their wisdom is a difficult design problem. From another point of view, the problem is not so difficult, we simply estimate the red sky before the "Happy Farm", how many netizens spent on that amount of time, reduced to the wisdom of time spent, this is a huge number! How exciting it would be if we had a certain intellectual task in such a popular game, and the user could help us solve the intellectual problem that one machine could not solve in the course of playing the game.

In fact, in the past days, it has been tried, as I mentioned above, that a class of games has its name game ith a purpose, that is, in the hope that through the game, people to solve some of the people themselves seem obvious, but the current machine learning methods still can not do the problem, For example, the image content recognition problem. So far, one of the most famous cases of this kind of thinking should be reCAPTCHA, a game that has successfully helped people solve problems that some of the contents of a printed matter are not recognizable when scanned into electronic objects. Such a human-driven computing thought, the domestic people translated as "human-oriented calculation."

This will be reserved for later discussion.

Source: http://www.wentrue.net/blog/?p=995