This accounts for the fourth paradigm of intelligent product recommendation recommended first official account. Accounts based on the computer field, especially in cutting-edge research related to artificial intelligence, it seeks to share more information about artificial intelligence knowledge to the public, to promote public understanding of artificial intelligence from a professional point of view; also hope to artificial intelligence personnel a discussion, communication, learning, open platform, so as soon as possible to let everyone enjoy the value of creating artificial intelligence.
Recently, the fourth held in 2019 recommended first paradigm "AI + new content" forum in Beijing, Yan strong deft vice president of business at the scene gave the keynote speech entitled "AI + DA short video drive efficient business growth", details deft AI applications in the process of commercialization.
Speaker Biography:
Yan strong, before Alibaba advanced algorithms experts, is responsible for the phone Taobao personalization algorithm; 2015 dual 11 personalized algorithm responsible for promoting personalization into the big promotion operations in the heart, as an electrical business big promotion activities as standard, creating electricity supplier personalized year.
2016 joined deft, deft responsible for the commercialization of research and development, build business cash deft technical system from 0-1. 2018 has been responsible for commercial business, dedicated to the use of AI and big data analytics to build the user experience and business needs the joint development of marketing system, deft exploration for commercial business model. 8 + years of big data, artificial intelligence, machine learning experience, for large data, artificial intelligence, industry trends and customer demand has a profound understanding.
The following lecture shorthand draft:
My name is Yan strong, the main business is in charge of deft AI by bringing value to our customers and short video community.
Deft established in 2011, at the end of May this year, it broke through 200 million DAU. There are currently 10 billion network-wide largest original short video inventory, adding 15 million short video content every day, all year-round use of the user's total duration adds up to 500 million years, the average daily number of video play 15 billion, 350 million daily like.
Commercialization has made a very important foundation is the traffic, our users what user do? The proportion of 90 after more than 80%, male to female ratio is 50:50.
In addition to being a deft high-traffic platform, it is a the "old iron Relations" featuring community. Has a natural credit system between the old iron, can produce very good relationship of trust.
In social relationships, we observed from the entire community ecology has emerged relatively good business ecosystem, so we quickly began to commercialize in 2018. We have a creator, he has 80 million fans last year, "two-eleven" He sold 160 million of goods. Relationship based on trust, he can very well do product recommendation.
Racer has a strong flow system, healthy system and trust in the value chain, based on traffic volume efficiencies and value to help customers stay well have to touch up the C-terminal user, for customers precipitation social assets.
How to enlarge the business deft ability to use AI technology? There are four main aspects: first, deft great emphasis on user experience, we took a long time to do a set of user experience as the core R & D system to guide the entire recommendation algorithm; the second, a precise delivery according to the needs of users; third, based on understanding of multimedia content for commercial content creation, production and commercialization of content through efficient way to assist the creation of AI; Finally, and most importantly, customer-facing needs help customers through intelligent operating system.
First, we attach importance to the value of the user's behavior. We calculate the precise value of each user's thumbs brought in the user system, in the process put in to link user interaction system of values and commercial value system.
Followed by accurate delivery. Based on the understanding of users, content, commerce, and the efficient matching of advertisements, in which the entire system will also consider various mechanisms, user experience and efficient mechanism to ensure accurate delivery.
The third is the intelligent operations. Advertiser's marketing needs and the way they operate determines what kind of matching advertising. This step we have a lot of manual operations involved. We put the whole process is divided into different stages, carried out to enhance the use of machine learning algorithms and at every stage, and for selecting different groups of people, should be different ways to achieve the best ROI. In the process of our business partners and are connected by intelligent operating system, put the whole process becomes more efficient. At the same time, we will optimize the details of each product operating in the process, create the best collaboration of man and machine status.
On the content creation business, we also have a lot of attempts. In deft, cover video is critical. For example, such an element of the cover, the effect of factors including its base map, copy, style, these parts ordered out of a number of combinations can be for different people, in different combinations to attract them.
In addition, through better data accumulation, have different adapter for different groups of people. After a picture has, by way of computer vision with AI adapter, and then meet the copy, with the launch of the process itself connected, this is a simple example on commercial content creation.
Last year, we launched a training system for advertising recommended GPU as the core. Currently, on a single machine to achieve a hundred T-level data and 2.5 billion the amount of training samples, training to enhance the efficiency of stand-alone 640 times, very effective to enhance overall efficiency.
We look at an attempt to help creators realized. There are a lot of deft creators, these creators include ordinary users, MCN. Before deft it has provided a lot of way to commercial realization, such as broadcast, electricity providers. This year we made a big change, in order to help content creators quickly realized the hands of better advertising model, we will understand the contents of the entire video after a user posted a video after the match to help users understand advertising. This year, we will strongly help the creators have better business for cash, so they have a better income.
Related reading:
Application AutoML in recommendation system
How AI enabling new media technologies fall?
To learn more, please search and follow the recommended first public micro-channel number (ID: dsfsxj).