So search official online Liu: differentiation to achieve subversion

So search founder Liu

Young Lions, Phantoms, lightning, MiG ... If it is not a military fan, it may be difficult to immediately realize the intrinsic relationship between these phrases--they represent, in fact, well-known fighter jets of all countries. And this is the name of all the internal conference rooms in search (yunyun.com), and it looks like it's carrying some kind of "fighter".

Differences

2010, from Google China web search research and development in charge of the Liu, led the company to lead more than 10 Google engineers, together to start the search business. Liu has always had an idea about search.

Google is not the first search engine, but by introducing the relationship between the Web page in the algorithm (PageRank), in the search field a revolution. In Liu's view, there is no large-scale improvement in the quality of search technology, the entire industry is basically in a state of maintenance. Search how to evolve, what will be the next generation of search engines?

"It's the Liu that needs to be added to the human element," he said. Relatively speaking, the search engine is not a light business, so that the previous two years and so search has been built in the basic search system, until the beginning of this year, and so the search to determine the way to the human-related "signal" and the search system combined.

Last night (December 17), after more than half a year, and so search official online.

Unlike a general search engine, a registered user (or five partner accounts) in such a search after login, everyone will get a different search results, and social integration of personalized results. Take the keyword, "Tae-So," for example, the more people (especially your focus) share on a page's social network, the higher the rankings.

Liu from four points of view: relevance, the introduction of social signals to improve credibility; coverage level, search results provide a richer range of content, the aging level, to solve the search engine for real-time content crawling slow pain, experience level, enhance the user's sense of search participation.

"Our timeliness is a order of magnitude", Liu Sina Technology said quality is the key to search, he laughs that he found a lot through the search for the original did not know things.

Trial and Error

From the idea to the reality, is a rough process. This is no exception to the search team. In the process of introducing social content into search, there are at least two most representative challenges.

The first problem is relevance. For such a search pattern, it is necessary to deal with large-scale real-time data from social networks and to correlate these real-time data with thousands of pages. "It takes a huge amount of data processing power," Liu says, and that the difficulty is not a small-scale real-time data processing comparable.

The second problem is the change in results. The number of people that a user focuses on on social networks, what they send out, all of them in unpredictable dynamic changes. And so search is needed from these dynamic content to find results, the common caching processing mode in this process completely ineffective.

"At first bad performance," Liu recalls that it took 5-6 seconds to return to the search results, and the team initially underestimated the amount of data to be processed--20% of the search results, and the content on social networks. To solve this problem, the algorithm and machine architecture should be reformed.

5-6 seconds, you need to shorten to 100~200 milliseconds.

One of the improved methods is to use flash memory instead of hard disk as storage medium. The random read and write speed of the hard disk is about 50~60 times per second, while the random processing speed of flash memory can reach twenty thousand or thirty thousand times per second.

"This is not the same as a traditional search engine," Liu admits, which does not constitute an absolute technical barrier to traditional search engines, "but see how fast you are and how determined you are to change the underlying architecture." In addition, he has left another protective weapon: patent, although it is difficult to comment on the effectiveness of this approach.

Subversion

So what's the competition for search? We're going the wrong way. "Liu in the interpretation of this not at the same time, and so the search does not want to enter the competition, in his view some of the traditional mode of search engine peers, although the mathematical formula to calculate the market share, but the ultimate success has been capped.

New exploration only has the possibility of subversion, Liu said to Sina technology, only differentiated competition, can provide unlimited imagination. He added: High risk makes a high return. However, there have been peers on different occasions or outspoken to say that the idea of the search is a gimmick, or show a pair of indifference unwilling to talk more.

However, no one can have a clear definition of the search engine tomorrow. Which is the way to the road? Or is there only one way to do this? No one can foresee the future.

Liu has its own set of theories. The Google PageRank technology mentioned above has also played a decisive role in enriching the Internet page, as it has the opportunity to show more pages. Similarly, Liu hopes that such searches will have the same complementary and mutually reinforcing relationships as social networks.

He concludes: "Big logic, for search engines, the more people prosper, I will be more valuable."

Social networks also need information integration, so search wants to do the job. The next search will also continue to explore the integration of more open social networks. While the industry has a view that search traffic is declining, Liu says how far the social network can go and how far the difference in search can go.

"I want to subvert the current search patterns and patterns," Liu The search for such a confident, "we will certainly encounter some squeeze, but not so afraid."

Wind say

Just days before the official launch of the search, there was a growing debate about the start-up company. There is even a claim that the founder of the search team within the Google Department has been in the financing after the basic set out to leave.

Google's team of entrepreneurs is one of the characteristics of such search. Sina Science and Technology This March report mentioned, and so the founding team from Google China members include Sun Yu, Chenli, An Xinghua, Wang Dong, winding, Xiang Kun, Li Dahai and other more than 10 engineers, as well as Cheng led the product design team.

Chenli and Cheng were pushed into the spotlight in rumors.

These two people have indeed been out of the search, will soon take over the Xinhua news agency and the mobile combined to search the CTO position. Cheng and other core team members have not changed. In addition, search for some of this year's business adjustments, such as abandoning the social networking plan, has also brought some people to leave.

"We do have some movement recently, mainly the poor performance, we need some elimination mechanism," Liu said that the so-called Google team to leave the argument is "disinformation slander." He told Sina Technology, from Google and so the search core team has been very united, and the spirit of the past is completely different: "In large companies, you do things quickly slow down, now every step is very solid, no retreat."

So how does search keep employees? One is the share incentive, but Liu says the IPO cannot be goal, otherwise will be quick success; The second is the ideal, Liu closed his eyes and thought for a while, and stressed the ideal of the word, he also said "we are really doing a relatively large innovation, a meaningful thing."

"The most important thing we've learned is to find a way to win," Liu said in a recent venture.