Archie Institute of Art Data: How to produce a film and television drama under large data

Under the big Data concept, the American video web site Netflix's 100 million-dollar series "card House" quickly became popular. This lets domestic video website to stir up.

Can the power of data guide the filming of homemade dramas?

Archie Art Data Research Institute Dean Ge Gengzhi told reporters, Archie Art is currently Choupai three network drama, and choose these three popular network novel theme, All is based on large data analysis.

On this basis, Archie will collect the user behavior generated on the website every day: including where users will pause, playback, fast-forward, etc., if a large number of viewers in a certain node to do fast forward or playback of the action, Archie can judge the user likes or hate the bridge section, and to guide the production of homemade drama.

In addition, Archie will also collect users on the site of the viewing behavior, according to these behaviors will users classify and "portrait", and accordingly targeted advertising.

Even so, Ge Gengzhi Frankly, the success of the network play and the theme itself, script and excellent production inseparable, can not overstate the impact of the data. Moreover in the actual operation, the domestic video website main profit pattern or the advertisement, the user pays the habit not to develop, this means that it is very difficult to let the user decide the movie and television drama actor, the director, the script completely. Another compartment, large data mining, modeling and analysis of the threshold, is still very high.

Two dimensions of large data

"21st century": For now, big data is a very hot concept. Archie What research and progress is there in large data?

Ge Gengzhi: Archie The study of large data is mainly two aspects.

One is the content of large data how to serve the user. For the user, our ultimate goal is to let the user see what he wants to see without having to pick what he wants from a bunch of content. We intelligently recommend what he is interested in by analyzing a person's viewing habits.

In addition, large data is also in the production of content to provide some help, traditional film and television in the production, more attention is the big subject itself and the script itself, including the use of the director, actors. To the era of the Internet, we can even use one of the plot or a variety show in a bridge section to analyze, the user's view of the plot is high or low, so as to draw the user's preferences to guide more detailed operations.

In addition to the latitude of the user, the other latitude of the large data is how to serve advertisers, that is, to help advertisers find the right audience, or to find his customers and potential consumers, and even to help customers find the consumers of their competitors, and marketing to consumers. For example, through the cooperation with Baidu, we can learn to watch video users in the past in Baidu search what content, so on the basis of advertising push.

"21st century": as we all know, Netflix's "card House" is a successful example of big data used on video sites. So in the homemade drama Hot Now, Archie Art also through large data analysis to guide the theme of homemade drama?

Ge Gengzhi: For the play of the House of Cards, Netflix has packaged it as a model for the success of a large number of data, with the core purpose of Netflix having to differentiate itself from the traditional film and television productions, such as HBO.

He needs to advertise his own features, which are characteristic of the internet's big data.

In fact, the most fascinating part of the play should be the subject itself and the script. To a certain extent, the success of the "card House" is the success of the subject matter and screenwriter, with large data, directors, actors and other relations are not particularly close, so we should objectively view the success of the card house and large data in the film and television creation of the role played.

Of course, big data can really help us analyze what subjects are of interest to users. Archie Art itself also uses large data to excavate themes. We have now turned on the three online dramas, which are actually internet-based data analysis. These three works come from well-known online novels, regardless of online reading or offline sales, the three novels are highly concerned. On this basis, we have decisively purchased the copyright and turned it into a TV show to move on the screen. In addition, which actors have a better reputation, actors and TV drama between the subject matter, we need to use large data analysis.

"21st century": in the video drama or variety show procurement, Archie Art is how to conduct data analysis?

Ge Gengzhi: In the film and television play procurement, we have a set of large data analysis process behind. Through the Archie of similar subjects, similar writers, similar directors, and similar actors, we speculate that the upcoming TV dramas may produce results in the future, thus assessing whether the play is worth buying.

A lot of TV dramas are on sale, but they haven't finished yet, even some of the more popular plays have not been filmed, only a script has been sold. As a video site, we definitely need to have a relatively accurate analysis and prediction, from this point of view, our historical data can help a lot of busy.

The value of the user "portrait"

"21st century": in the advertising push, how do you through the data analysis to carry out the crowd positioning and "portrait" of?

Ge Gengzhi: In advertising push, we have developed a lot of products in the past two years. To give a simple example, if you have searched the BMW on Baidu in the last one months, when you come to Archie art to look at any content, i know you have searched the BMW car, I can give you the advertisement of BMW, of course also can launch the advertisement of Mercedes Benz. This is the core value of the product.

We have also developed a product called "groupies" this year, as we all know, many stars have their own fans, such as Chao fans will watch Chao related ads, will also watch Chao TV dramas and movies, and may even look at Deng-related variety shows. When we capture the user's multiple viewing behavior, we define him as a Chao fan. Then, we will give him the ads by Chao endorsement.

Generally speaking, the user's information is divided into two categories, one is the user's natural data, such as gender, age, region, etc. the other is his behavior data on the Internet, including his search behavior, viewing behavior and so on. We think that user behavior data is more important than his natural data.

"21st century": So, Archie currently divides the user into several kinds of categories, or how many kinds of labels for users?

Ge Gengzhi: There will definitely be hundreds of tags, because there are different levels. For example, according to Baidu's search data, we can give him a label that he likes different kinds of consumer goods. For example, the person like the car, the person likes health care, another person like beautiful skin care, which is based on his interests and focus on the field of a label, such a label may have dozens of, or even hundreds.

Another kind of label is about what types of movies and TV dramas users like to watch. Some users like the subject of gun battles, some users like American drama, and users like love movies and so on. There may be dozens of more tags in this series.

In addition, there are tags related to the user chasing stars, such as this person likes Chao, that person likes honglei. These tags are divided into different dimensions, each with dozens of or even hundreds of tags. Some users can post five or six kinds of labels at the same time, that is to say, he is fit for five or six different kinds of ads.

"21st century": On the basis of user classification, how will advertisers generally choose audiences for delivery?

Ge Gengzhi: Generally speaking, if it is cosmetic day, food and beverage and other consumer goods industry, advertisers will not only pick a category of users, but will pick several types of labels users. But if you are a high-end brand, or a product brand for a specific group of people, such as you sell the server, then your audience is certainly not ordinary people. Can have a server procurement requirements, may also be so tens of thousands of people, hundreds of thousands of people. At this point, advertisers need to add a few categories of tags, to find the overlap part of the people, these users will be very valuable. Therefore, how to put the advertising and brand in the industry and his audience is closely related to the scope.

The big Data view of Youku's defection to Ali

"21st Century": Archie's current large data analysis method, do you think the accuracy rate is high?

Ge Gengzhi: This can't be generalized. For example, in terms of program procurement, the history of the flow of data to infer the heat of domestic TV dramas, the current accuracy can probably reach more than 80%. The regularity of this data analysis is relatively strong, so the accuracy is relatively high.

For some overseas dramas, we are more likely to look at its broadcasts abroad, mainly in ratings and in the spread of social media abroad. It is possible to forecast domestic broadcasts through overseas broadcasts, but because of the different cultures of the regions, 30% of them may be unexpected. As we broadcast earlier this year, "You from the Stars," the Korean drama, its broadcast in South Korea, the heat is far from high in the country.

Variety shows are not the same as TV dramas. Because of these annual variety shows constantly introduce new, hot switch very fast. The first two years of the fire is the singing talent show, the beginning of last year is a parent-child class program, this year has become a star reality show class show. This new subject is more testing our analysis of the data system, because these topics have not appeared, no historical data accumulation, analysis of this has a certain degree of difficulty. We will also make forecasts by referring to the broadcast of similar programmes overseas. We continue to accumulate experience in this area.

"21st century": What do you think is the main challenge for video sites in the mining and application of large data?

Ge Gengzhi: The big challenge now is that video sites can cover a relatively limited range of user behavior. Or, the video site mainly covers the user's leisure time. Then the user in the recreation, his work, shopping, his hobbies, consumption habits, as a video site is not available.

That's why Archie is a subsidiary of Baidu, and Ali shares the Youku.

You will find these video sites, more or less related to bat. Because from the perspective of the video site itself, we also need more level of user data. In other words, the core purpose of large data is to depict the user, when we depict a person you can not only understand one aspect of him, you must know him in every way. This is a video site on the big data challenges, now everyone is trying to solve this problem.