As the next stage of the “background front-end†of the journalism in the digital environment – ​​the main body of the “post-shiftâ€, the news push service of the platform media (plaTIsher) is mostly based on the user's usage habits, and the algorithm predicts the user's interest. Information and topics are pushed to the user. The social media that gradually transforms into platform media dominates the aggregation and distribution of news content in the existing mode. The main positioning of its identity and function is through the accurate matching of user preferences, from the massive information flood. Select the information that the user cares most and is most suitable for the user to receive. This model is currently highly regarded.
The US Pew Research Center's 2016 US Social Media Platform News Usage Report shows that 62% of American adults get news from social media, and Facebook becomes the largest news portal; China Mobile's information distribution released by Analysys in August this year The Market Research Report 2016 shows that in the domestic information and information distribution market, the content of algorithm push has exceeded 50%; the Reuters News Research Institute survey shows that although there is indeed an Algorithmic Censorship and platform bias for algorithm push news ( Platform bias), informaTIon cocoons, and echo chamber concerns, but people, especially young people, are more willing to use and believe in content generated and pushed by algorithms.
For the media industry, there is probably no era that emphasizes the role of technology more than today.
The major information platforms are fully committed to the algorithm, and I hope to share a piece of the wisdom media era. Whether it is the headline of the $1 billion D-round financing in April, or the first three days of the news, it is the first three newsletters of the information application, or the new information of the respective information, Netease, Baijia, etc. The products all emphasize their own technical attributes and channel distribution advantages. In front of the algorithm, the content is king's voice seems to be much smaller.
In the game of giants competing to build ecology, high-quality self-media and professional media have become the object of competition, but they also face choices: cooperation with the platform, collusion with algorithms, or continue to take care of the fact that “the wine is not afraid of the alley†face? From the actual situation, there are many similarities between domestic and foreign countries. Most producers and production organizations choose to find their own comfort zone in the new ecology, which means they have to accept this ecosystem. The law of the jungle.
What is a recommendation algorithm?The recommendation algorithm is an algorithm in the computer profession that uses some mathematical algorithms to infer what the user might like. The recommendation algorithms are mainly divided into six types: based on content recommendation, collaborative filtering recommendation, rule-based recommendation, utility-based recommendation, knowledge-based recommendation, and combination recommendation. This paper selects two common introductions:
1 content-based recommendation algorithm
The principle is that the item is similar to the item that the user likes and the item that he has paid attention to. For example, you read a news about the “Chengdu purchase restriction policyâ€, and the content-based recommendation algorithm finds a piece about “the price of these new discs is very low after the purchase restrictionâ€. The news that the high-tech zone is expected to be less than 10,000" is closely related to the news you have watched before (there are many keywords), and the latter is recommended to you.
2 collaborative filtering algorithm
The principle is that users like news that users with similar interests like, such as your friends or people you care about like sports news, then it will recommend it to you, this is the simplest user-based collaborative filtering algorithm (user -based collaboraTIve filtering), and an item-based collaboraTIve filtering method, which reads all the user's data into memory and becomes a Memory-based Collaborative. Filtering, the other is Model-based collaborative filtering, including Aspect Model, pLSA, LDA, clustering, SVD, Matrix Factorization, etc. This method has a long training process, but after the training is completed, the recommendation process is faster.
There are three strategies for general news recommendation:
1 content-based recommendations
It can also be called a recommendation based on the user's portrait, which means that the user clicks on the record according to the history of the user, summarizes the user's preference, that is, the user's portrait, calculates the similarity between each news and the user's portrait, and recommends the news with the highest similarity to the user;
2 collaborative filtering recommendation
Is to find a group similar to the user's interests, and then recommend this group to the user;
3 popular recommendations
Another common method of news recommendation is the popular recommendation. This is to set a time window to count the traffic of all news in the past period of time, and recommend the news with the largest amount of clicks to the user;
CIXI MEMBRANE SWITCH FACTORY , https://www.cnjunma.com