258. Feeling that IQ is crushed

Cao Yang looked at the confused people and smiled slightly, "I call this thing - accurate push."

"Accurate push?" Everyone lowered their heads and thought.

As the Internet becomes more and more developed, more and more people are surfing the Internet. The information in this world is simply an explosive growth, and everything will be presented on the Internet. Every day, everyone is actually the maker of information, let alone various large enterprises and big countries...

Celebrities have endless melons every day, news happens every day, and every spiritual Yue has something fresh every day...

When these things are flooded on the Internet, an era of big data has actually ushered in.

The word big data will not become popular until 2010. People use it to describe and define massive data in the era of information explosion and name related technological developments and innovations.

The first one to do [precise push] will be in the headlines in 2012.

It is really not easy for Toutiao to become a company that can compete with Penguin by relying on intelligent algorithms related to accurate push.

It also shows how powerful the precise push is.

"Yes," Cao Yang said seriously, "In fact, in the era of data explosion, everyone's attention is very limited. Most of the information is redundant information for people. They don't need this. Everyone's time is limited. They only pay attention to what they like to pay attention to.

What we need to do is to analyze and extract users’ hobbies, and finally push them what they like.”

"Developed an accurate push algorithm, and this model can be applied to various fields of our company. For example, as mentioned earlier, which type of game players like to play, the relevant games will be pushed to them, and what type of songs they like to listen to, and then this type of song will be recommended to them more.

In Singularity Novel, I suggest adding a push page separately so that players can see what they like in time. At the same time, when the chapters are updated, they will be displayed first in the recommended types he likes."

This is what Cao Yang insisted on, not repeating the wheels.

A set of artificial intelligence algorithms can be applied to various industries, which is actually a huge profit.

Toutiao did the same back then.

Alibaba repeatedly created two identical models and systems, which resulted in a lot of waste of resources. Later, they also transformed into a middle-end.

It is actually not cost-effective to make a separate set of adaptations for each product.

Since Cao Yang is a veteran, he must know this better than everyone here.

"Well..." Everyone present fell into deep thought.

The two programmer bosses had an ominous premonition.

"It's great. I support it." Liu Zhiping looked up and applauded.

The others were stunned for a moment and applauded and agreed.

Because from the proposal, there is no reason not to support it.

Accurate push actually helps improve product push efficiency. At the same time, users can find what they want in a short time, saving a lot of time redundancy.

Time is money, this is true for users, and even more so for Bomi Company.

At the same time, precise pushing can give users what they want, and can enhance their corporate image, making users more dependent on Bomi's products, thus opening up the gap with similar products.

This idea is simply a genius.

...

"But, how to do this?" Wu Wenhui asked a question.

"It depends on our program bosses." Cao Yang said with a smile.

Liu Guangran and Deng Peiqiang's chrysanthemums tightened.

Sure enough, what should come is still coming. As long as Boss Cao holds a meeting, the program will inevitably suffer in the end.

Liu Guangran looked confused.

"This is not my learning scope. Is it necessary to use artificial intelligence algorithms?"

"Look for someone." Cao Yang said.

"And, this thing is complicated and simple," said Boss Cao, "In principle, what we need to do is label the data and also define the user's portrait.

For example, big data analysis, accurately classify customers, and can be divided into: ageing, regionalization, gender, hobbies, etc. for precise distinction. In this way, as long as the user uses Bomi's products, the platform will constantly "steal" the user's related interests and hobbies and push them accurately.

Then our data will be labeled with various tags, and at the same time, a set of weight value algorithms need to be given to increase or decrease the user's [Perference Weight]. It is better to say that every time a reader clicks on an urban novel, we will label him an urban tag, and the weight of the city will be increased. This is the reader's preference weight. This weight will decay with time, and there will also be a corresponding threshold.

When a user clicks on a fantasy novel in the recent time, the preference weight of the fantasy novel will be increased accordingly. When we promote books, we will provide the corresponding books because of the user's preference weight.

The reason for decay over time is that the user's taste may not be constant, but they will remain stable within a certain time frame.

The more detailed our data labels are, the more refined the classification difference and overlap, the more accurate our push will be.”

Liu Guangran thought silently, "Well... this seems to be feasible."

Deng Peiqiang added, "But a very core problem here is how to divide the granularity and then disassemble the data. We definitely don't want players to end up like to eat tomato and eggs, right?

Instead, I hope that the player will withdraw from boiled eggs, egg noodles or cakes on the basis of eating tomato and eggs?

These things actually have high requirements for data screening.”

"Well," Cao Yang agreed with Deng Peiqiang's words. There needs to be a smarter judgment standard between precision and fuzzy. "Then the difficulty lies in this set of algorithms, about how to analyze users' hobbies and how to distribute data, this is a test of you."

In fact, after the large algorithm is defined, the remaining parameters can be adjusted completely through the actual data situation, and in the end, it can even achieve perfect automatic adjustment, allowing the entire algorithm to automatically adapt.

However, what Boss Cao said just now has given two program bosses a great reminder.

Toutiao's algorithm is also improved and corrected step by step in practice. As long as it can start, it can be improved slowly.

Cao Yang believes they can do this.

"You can try it." Liu Guangran said.

...

"Does anyone else have any problems?" Cao Yang saw that the program could be handled, so basically he would be handled. A set of accurate push algorithms will run through the entire Bomi system in the future. After the precise push is released, the user's adhesion and dependence on the Bomi platform will be greatly improved.

The sooner you do this, the sooner Bomi will get massive data and experience.

To put it bluntly, whoever starts first will lead.

ByteDance has done this first in the industry, so they have a lot of data and experience to correct their algorithms. This is polished by time, so ByteDance's products have always been ahead in pushing, and even Penguin has not caught up.

Everyone was thinking carefully.

After a while, the new Wu Wenhui raised his hand and asked:

"What should I do with my original ranking? With intelligent push, will our editorial power be reduced? How to deal with the official promotion power?"

Cao Yang smiled slightly.

This question is well asked.

The contradiction between this problem lies in the contradiction between [centralization] and [decentralization].

Because before the emergence of accurate push or intelligent push, the exposure of the book on Singularity was completely controlled by the official, and when sent to specific people, it was actually the editor.

It is up to the editor to decide which book to put forward what kind of recommendation position to put in next week.

These recommended positions are planned based on the eye-catching or exposure of the website. For example, at the top of the homepage, the recommended position given is the largest and best page. This recommended position is called a large push, and it will generally give works with very good grades.

At the same time, the books on the ranking list will also be ranked according to the performance and popularity of the books.

How to give recommendation positions and increase the exposure of works is actually controlled by the editor, or in the hands of the website.

This is centralization—

The official will decide what the book is exposed.

The official has a high voice.

Once intelligent push appears, it is actually equivalent to [decentralization]. Users can directly refresh the content recommended by smartly and click on the book they like to read, and then they will continue to get what they want.

Maybe I won't read the editor's recommendations.

Decentralization will definitely have a certain impact on official power, which is equivalent to the transfer of part of the official power to the exposure of works to intelligent push.

"But, what does this matter?" Cao Yang said, "I am also worried that the official power will be too great and delay a lot of good books."

"Ha?" Wu Wenhui didn't understand very much.

"Actually, doing this will mean that every reader sees different web content. This is actually a good thing. Think about it," Cao Yang said with a smile, "The different web content of every reader sees different web content, does it mean that we can provide more recommendations?"

"Ha?" Wu Wenhui still didn't understand.

"I'll tell you this."

Cao Yang picked up a marker pen and drew the structure of a website on the whiteboard. "Suppose the recommendation positions of our website now include big pushes, Sanjiang pushes, and other types of recommendations, which are not important.

Think about it, how many recommended positions can we give in a week for a big push?”

Wu Wenhui replied with a confused look on his face, "There are 4 recommended positions a week, plus the rotation every two days, we can recommend 7 large blocked items in a week."

"Well, then for readers," Cao Yang pointed to one of the empty spaces, "I like to watch fantasy types, or add another one. I like to watch fantasy and Western fantasy types, but I'm not interested in the city.

The effect of your recommendation position is actually discounted.”

"How could that happen?" Wu Wenhui said in confusion, "I am pushing it to the entire website, which is equivalent to seeing it by all website users. Even if the conversion rate is low, at least the recommended data is still guaranteed."

"Actually not," Cao Yang drew several circles and let them overlap to a certain extent. "I admit that user attributes overlap. Some users like fantasy and cities.

But there are also users who are different, they just don’t like the city.
To be continued...