Big Data help Custom Love

Otaku are always fantasizing about customizing a girlfriend for themselves, perhaps their wishes will soon be reached. Instead of using the 3D printer, it relies on dating sites in the big Data age.

At the end of 2012, NetEase's new dating website, "flower Field", was online. Hadanda to free communication as a selling point, the main one or two-line city high-end market. Abandoned the traditional marriage website "Artificial Matchmaker", the flower field does not provide any artificial service, from the recommendation to search all by the system automatically completed. And the industry's leading boss century Jiayuan (Nasdaq:date), also in the hot after the fight, trying to use large data precision operation, in the mass of heterosexual data in the search for the user "refers to a Ming road."

A year ago the love website most likes shouting "the real name". Through the user ID card or mobile phone number verification, the site can obtain the user's real name, age, gender, area and other identity card data, to prevent fraud has a certain role. But now all the families have, this traditional play is too angry. After all, people with identity cards are not necessarily people with identities. Marriage website has set the attributes list complete, people almost suspected of entering Taobao meat market. When the height, weight, age, education and so on are all framed, see there are 2000 girls (boys) meet the requirements, it is time for rational return to sensibility. And this sensibility, is based on the rational data analysis.

The Flower field interface is fresh and concise, using a kind of micro-blog information flow display form. Home page is a recent update of the opposite sex users of the picture, inner monologue and text to show their lifestyle, personal taste and other soft materials. The system automatically recommends people who are relatively active and have recently updated the flow of information, prompting users to refresh their tweets to update the fields and accumulate a large amount of soft data to analyze.

"We can use the ' inner monologue ' to tap into the character of the user. NetEase Flower field director Xiatianyu said. Xiatianyu has been working in NetEase for 11 years, before his team was responsible for the development of NetEase Weibo, and his earliest job was Oracle database management.

Now, he and his team are trying to decode the user's personality through natural language processing and semantic analysis to achieve a "carrot and stick" recommendation. First, they use the word segmentation method, from the user's "inner monologue" to extract the high frequency of keywords, and then classify these keywords, such as perceptual vocabulary or rational vocabulary, finally, through the text analysis, semantic analysis, from the user's personality is introverted, extroverted, idealistic or realistic pie and so on. This technology is still being tested, and will be put into use gradually in the second half of 2013.

Since December 28, 2012 to open the National Register, the flower field registered user has nearly 250,000, daily active user 40,000 people. Although the development is rapid, but compared with the industry predecessors, the flower field users are even less attractive to cheat. But NetEase's technical reserve is not to be underestimated. "Youdao" search provides natural language processing techniques that are used to excavate user personalities through text. NetEase was originally used for NetEase mailbox face recognition technology, but also the love and marriage site can be a powerful tool.

The Flower field team only has more than 30 people, mostly 85. On the basis of analyzing the massive soft and hard data, the author sums up some characters and sets up a certain number of character models. Then analyze the specific users and classify them into various models. In this way, the user will be able to recommend this type of person to one of them.

This model is not only a character model, but also a model of appearance. "We're going to launch face recognition right away," he said. For example, you want to find a girl like Fan Bingbing, you enter Fan Bingbing, will recommend to you a lot of fan bingbing face of girls. Xiatianyu introduced, flower field backstage has extracted fan bingbing face of data, will also launch dozens of kinds of popular male and female star face for users to choose.

More interesting is that if the flower field can cross the product platform, combined with NetEase portals, mailboxes, games and other NetEase resources for large data analysis, whether it can recommend to users with their own reading habits, work habits, entertainment habits are matched objects? The real big data must be cross-platform, and it's hard to do that elsewhere.

But the cross platform data extraction to some extent also involves the user privacy. Where is the bottom line for big data? Xiatianyu that the data involved in the bank must not be touched. This is a very wide range. However, it can be determined that NetEase will not be in the field of influence and users are not enough time to open cross-platform data analysis.

"I hope Hadanda can build a five-storey building with different categories of people on each floor," he said. A person who earns 50,000 bucks comes in and doesn't feel like there's a lot of cock in here. Xiatianyu describes his vision for Hadanda. Although the Prince with Cinderella's romantic fairy tale, but also real for users save a lot of screening trouble.

Data analysis can not only be used for accurate recommendation, but also to identify the most criticized dating sites fraud and fraud. At the end of 2012, founder Gong resigned as the joint CEO of Century Jia, another joint CEO Wulin Light as CEO. Wulin in the air network as Vice President and mobile gaming Division general manager. After joining the century, he first made the introduction of large data to improve the user experience. Century good edge of the data analysis team developed a network of police system, from the previous passive waiting for users to report fraud, instead of active attack.

"For us, flower basket childcare and bar care are the most harmful." "Century Jia Margin CEO Wulin Light Introduction, these two kinds of people are characterized by the hope of long-term operation, not like real fraudsters, so that victims report." The money is not much, but the most disgusting.

Bar care is characterized by beautiful photos, but only one or two. The flower Basket Child care is the good-looking, the speech gentle middle-aged uncle, but only calls does not meet. One day he claimed that his company was going to open, and asked the woman to send a flower basket and cheat one thousand or two thousand pieces. Their common feature is to contact each other as soon as possible, leaving the network platform, using the telephone this difficult to control the way to establish intimate relations.

Through the data analysis of these behavior patterns of swindlers, Century Jiayuan made a set of cheater identification model, which has been put into use. The hilarious is, Wulin light I also because of testing products, was this set of network police system caught, "I meet the characteristics of the bad guys, come up to find women directly to contact the way." ”

Wulin the light to share a series of data with us. For example, most of the "leftover" was born from 1978 to 1983.

"Because of family planning, this segment of the population is soaring, their corresponding 1973-1978 male population is relatively insufficient, and their age of men were 85 after the young girl robbed." Wulin said that the 2009, 2010 Love websites broke out just because the people got to the age of marriage.

Another example, Century Jiayuan is trying to travel through the tour, the more in the century good margin of the people who do not receive the letter, playing more pages.

Hundreds of years ago, the matchmaker never imagined that a few years later, a group of skilled engineers with cold data for the YUANNV Baomei tow.

However, although technical men can identify the likes and dislikes, grasp the rogue, but also have difficult technical problems, such as face recognition. The judgment of the appearance is sensitive and personal, it is always the difficulty of the marriage website. Century Jiayuan has been trying to find suitable facial recognition software, but it is very difficult. Wulin Light says they are introducing a system that can identify similar faces, summarizing the user's preferred face shape from past selection data and recommending people with similar faces.

Hadanda, although backed by NetEase tree, is about to launch the face recognition function, but this technology is currently stalled in face recognition. Hadanda try to make more advanced face recognition, such as facial features recognition, couples matching as their own value-added services charge points, still need further technical breakthroughs.

To what extent can the data be achieved for such a complex animal as humans? Wulin Light said that their ideal is to allow users a registered century good margin, see the recommendations of the opposite sex are all like. But from all the complaints that are heard around you, it will take time to achieve this goal.