Explore the content granularity, and require bionic

Based on the subscription mode of grabbing shrimps, I have seen sayonly's article titled "granularity of excavation content", which is a way for people to read.
This method is natural, because we often listen to a series of lectures from a teacher and talk about the whole world for thousands of years. It will read all the works of a writer, which is the prototype of subscription by person.
Of course, it is not the only one.
Our most common mode is actually topic-based subscription.
That is, when we watch TV, watch magazines, read newspapers, and see a news event or perspective, we find it interesting. So we read the article and if there are any follow-up reports on the next day, we will also give priority to reading.
This is the topic reading method.

The granularity of the explored content needs to be bionic. Imitate user behavior, capture shrimp, googlereader, and Bloglines imitate reading by people.
So what about reading by topic?

There is also a natural way to read. My friend knows me, and the books he is reading in his small circle will be recommended to me first. This may be the "Sn reading method" described by sayonly ".
Now there are two ways to do this.
The first type is Douban.
Second, hard computing.
Our team has tried this experiment. We can recommend some blogs you may be interested in.Article.
However, the machine cannot do anything within two years. So far, I have not figured out how to combine the divergence of people with the rigorous computing of machines.

Assuming that the word "Mining granularity" refers to three meanings:
Discover people you may be interested in reading;
Discover topics that you may be interested in. [You may know this topic, or you may not know its existence, such as new hot spots, such as long-term topics]
Discover articles you may be interested in.

So we made a little interesting attempt to discover theme you might be interested in and explore the article you might be interested in reading: http://www.onejoo.com /.
From the perspective of the page, onejoo seems immature, but it integrates the following technologies:
A. the hot spots of playings are automatically discovered by machines. The Internet hot spots of the day are obtained through large-scale Text Mining and computing on a regular basis every day.
B. The vast majority of articles on every topic of playings come from machine clustering computing or from hot spots.
C. For each topic, the machine calculates subsequent tracing reports every day based on semantics.
Http://www.onejoo.com/TopicMix.asp? Iden = 1 & topic id = 840 [ What is the "trap" that the driver cannot hide when the speed limit sign is used "? ], Http://www.onejoo.com/TopicMix.asp? Iden = 1 & topic id = 826 [ Shanxi has set up millions of prize winners ], Http://www.onejoo.com/TopicMix.asp? Iden = 1 & topic id = 825 [ The number of professional beggars increases. Experts suggest issuing credit certificates for begging" ], http://www.onejoo.com/TopicMix.asp? Iden = 1 & topicid = 836 [ Li Yuchun grabs Liu Xiang's Limelight collective protest in the sports industry? ] The "time context" module of these topics is clearly displayed. " Li Yuchun grabs the limelight of Liu Xiang " And so on.

Technologies used for playmates:
New hotspot discovery computing;
Time context computing;
Clustering computing;
Crawlers are talking about cool news today if crawlers are still doing technical work.

This is actually the prototype of reading by topic.
We are planning another form of reading by topic to better explore the content, especially the personalized needs of Long Tail users.

Additional reading:

Technical hero meeting [2]: QCsdnAn information filter problem

Technical hero meeting [4]: also talks about how to discover the desired content and heroes

Principle of meme engine (I,II,3.)

The future andWeb3.0Of WeiCome