Current location - Loan Platform Complete Network - Big data management - Youku algorithm expert reveals: Yi Yi Qianxi, Lei Jiayin in "Chang'an" are big data casting
Youku algorithm expert reveals: Yi Yi Qianxi, Lei Jiayin in "Chang'an" are big data casting

Big Data found Lei Jiayin through labels such as Big Head and Tough Guy

Shen Yi: We have a casting model, and the dimensions are based on the Big Data refinement. There are two main dimensions: the actor himself and the performance of historical works. Through the whole network of public opinion data to extract the netizen's evaluation of the actor's label, calculate the correlation between the actor and the candidate (character) label; in addition, we also have to take into account the actor's current popularity and commercial value, including the face value. In addition, what he has acted in, what is his role in the works, how is the performance of historical works, how is the performance of movie box office, TV series traffic and rating, these dimensions are disassembled to the actor, and the combination of the above mentioned factors is added to match the needs of the business side. We generate a list of candidates and then choose from that list.

Shi Xiaofei: Casting ability is one of the application scenarios of the Beidouxing product, Beidouxing is a pan-content AI platform (pan-content refers to the content of the entertainment industry, which includes episodes, movies, variety shows, comics, artists, creators, etc.). Relying on AI+Big Data capability, Big Dipper analyzes pan-content in full-dimensional deconstruction, and gives AI+Big Data auxiliary decision-making reference in the whole life cycle of content investment, production, operation and marketing. The platform technology has been accumulated for quite a long time, and the product was exported about two years ago, before it was more used internally, and now it has exploded on "Chang'an".

Shen Yi: Zhang Xiaojing's character labels include tough guy, lone hero, smart, sense of justice, and then in the original, he has a nickname Zhang Daotou because of his big head, and Lei Jiayin's final matching labels include "big head," because that's what a lot of netizens mentioned in their comments about Lei Jiayin. This is what many netizens mentioned in their comments about Lei Jiayin. In addition, he has acted in anti-Japanese dramas and also has the image of a tough guy, and the labels identified by public opinion include tough guy, righteousness, handsome, strict, and so on.

Shen Yi: We have the face value model, which is calculated by machine learning. The principle is that there are a large number of real people who rate the photos of actors and stars they see, and then feed this data to the machine, which learns the logic of the people's scoring. This way the machine knows how to score and how many points to score when it sees a new photo. With this model we scored actors across the web, and Lei Jiayin ranked in the TOP 10, beating 90% of male artists.

"Change to two a week is more the advice of big data"

Shi Xiaofei: YiYiChangXi also used big data, in the casting of this decision-making process, the North Star will give AI + big data to assist decision-making Suggestions, the final decision maker is still the casting director. They will then make a final decision based on their own industry judgment, plus the actor's schedule, as well as cost-effective comprehensive consideration. ai + big data + director's industry experience to form a very good complementary advantages.

Shi Xiaofei: This is the North Star system to give advice, the user pop-up feedback and ratings data performance said that must be added, so from one week to two a week. This was adopted at the project meeting, and is the decision-making advice given by AI+Big Data.

Nandu: What are the pain points that are more difficult to solve manually that big data can help with casting?

Shen Yi: AI plays a supporting role, and the results of big data are more objective, which can avoid too many subjective factors, such as the degree of personal preference for the actor, the support of big data makes each dimension interpretable. At the same time, it can provide a broader vision of casting, the results of the feedback from the entire network of actors library, may be more comprehensive than a single casting director in a short period of time to think of candidates, to avoid omissions.

"The deconstruction and quantification of content is more difficult than commodities"

Shi Xiaofei: The data level support at the broadcast stage will be more direct than the casting, and the broadcast stage can be directly accessed to the broadcast data on the platform, AI Big data penetration at the platform will be much earlier than the content itself.

On a platform like Youku, all the viewing data, user interaction data, and direct performance of the content can be monitored by big data. We have AI big data to deconstruct and analyze the content from the IP stage to the script stage to the finished film stage, and try to quantify the quantifiable indicators of the content.

AI big data on the one hand from the platform side of the depth of understanding of the user, this piece of Tao system products on the user and platform content understanding is very mature, we are now through. From the content side, we then go to do mining and deep analysis of the content, the combination of the two is to explore how to do a better integration of content and platforms to play a greater value.

Shi Xiaofei: The deconstruction and quantification of content is more difficult than the quantification of the structure of goods. For example, e-commerce products 38 yards women's sandals, which quantitative indicators are very direct. But how to quantify the content "12 Hours of Chang'an"? This is relatively more difficult. For example, the face value quantitative indicators, we need to train a massive amount of data, on the basis of which a quantitative indicator is trained; the user side of the point of view, e-commerce commodities have a clear user, that Chang'an who will watch This requires a deeper understanding of the content consumption user. When we define content users, we can not simply set the gender age such basic dimensions, but also to establish psychological labels such depth indicators.

We will also continue to do product upgrades, such as the North Star will be from the script analysis to shooting production and other aspects of the link to provide from casting to the expression of the persona, the production period of performance, the latter part of the editing of a more detailed analysis of the proposal; will continue to promote and the promotion of play through, to provide a more accurate user guidance and reach the promotion and operation.