Author | Yun Zhao
Between users and information, there is either a search or A recommendation across the board. As Baidu Executive Vice President Shen Dou said at a conference: People are so familiar with search that they cannot feel the technological changes.
Today, search is everywhere, from browsers, WeChat, Alipay, to other apps that we log in to and use every day. We are accustomed to using search to filter the information we need. . "Search" has become a basic technology in the Internet era. It no longer has a "sense of presence" in our sights like new technologies such as blockchain and Web3.
What is really important is often what we take for granted but cannot perceive.
In the era of big data where the amount of information is exploding, traditional search has also evolved into the era of intelligent search. With a search engine powered by AI, the machine can not only read text, but also understand speech and pictures. AI technology provides the driving force for continuous evolution of search.
So, in order to make searches faster, more accurate, more personalized, and more intelligent in understanding user needs, what active innovations and explorations has the Internet made? Here are some:
In 2016, Google launched the RankBrain algorithm based on backpropagation technology, which can help users search for unpopular search results faster based on semantic analysis and vocabulary association. .
In 2021, Google will deploy BERT, the pre-trained language model that is popular in the NLP field, into Google search. Even if you enter a large paragraph of text, the search engine can get what the user wants to search. The introduction of BERT has improved at least 10% of search results.
As for domestic search, the application of AI in search is also in full swing.
In 2015, Baidu proposed the concept of multi-modal search, exploring the evolution from text search to multi-modal search such as voice, vision, and video; Tencent is pursuing vertical search The search field continues to develop, and technologies such as vector retrieval, heterogeneous computing, knowledge graphs, and video understanding have been successfully applied in Tencent Video and Tencent Kandian; Meituan Search has transformed into an AI search engine after years of technological accumulation. , which greatly improves the performance of business indicators in core search scenarios such as merchants, takeaways, and content.
#If search is to help people find accurate content, then recommendation is to accurately push content to the right people. In 2021, Kuaishou and Tsinghua University proposed a new sequence recommendation framework SURGE based on graph neural networks, which greatly reduced the difficulty of capturing users' true interests; in 2021, Alimama's algorithm engineering team and Alimama's external advertising algorithm team open sourced Elastic-Federated- The Learning-Solution (elastic federated learning solution) project introduces federated learning theory into Alimama’s business scenarios, taking advantage of its privacy protection and algorithm theory.
With the continuous development of AI technology, major Internet companies are increasing their investment in intelligent search and recommendation, a high-tech track that is closely related to users! If you want to unlock more cutting-edge technologies related to intelligent search and recommendation in advance, I believe the AISummit "AI-Driven Search and Recommendation" special session can give you the answer!
The wave of digital transformation has given rise to new evolutions in search recommendation technology, such as: traditional search engines are upgraded to AI engines, The focus of search has also shifted from general search to refined vertical search, and recommendation technology has further ushered in a stage of deep integration with intelligent algorithms.
From August 6th to 7th, 2022, the AISummit Global Artificial Intelligence Technology Conference will be held as scheduled as an online live broadcast on the official website of the conference, with an estimated 100,000 attendees. With the theme of "Drive·Innovation·Digital Intelligence", this conference is mainly aimed at mid-to-high-end technology managers and technical practitioners of technology companies, business managers planning/currently undergoing digital transformation, and people interested in the field of artificial intelligence. and entrepreneurs. The conference will also invite nearly a hundred technical elites from well-known Internet technology companies, managers of traditional companies in the digital-intelligence transformation period, and experts and scholars from cutting-edge academic institutions to jointly discuss the industry driving forces of artificial intelligence and discuss cutting-edge innovations in artificial intelligence. Technology, let’s talk about the wave of “digital intelligence” in the era of artificial intelligence.
At this AISummit conference, in the "AI-driven search and recommendation" special session, senior technical leaders and algorithm experts from Alibaba, Tencent, Meituan, Kuaishou and other industries will From the perspective of business practice, share forward-looking thinking in the field of intelligent search and recommendation.
With the continuous development of Meituan’s retail category business, Meituan Search’s related technologies in commodity business are also constantly iterating, and the sorting module serves as an important component of the entire search system. section, which greatly affects the user’s final search experience. In recent years, deep learning has been widely used in the field of ranking.
In this topic, Chen Sheng, the person in charge of Meituan Search and Sorting, will introduce the technical architecture and sorting platform of Meituan Search in detail with the theme of "Construction and Practice of Meituan Search and Sorting Platform" Every detail of the implementation construction and optimization of the sorting algorithm is discussed, and relevant technical experience is shared through practical cases.
Vertical search engines, unlike traditional searches, do not provide users with hundreds or even tens of millions of search results, but extremely narrow and precise specific information. Therefore, users in specific scenarios prefer vertical search engines, which is an inevitable trend in the segmentation of the search engine industry. How to find the real needs of users from massive information and user interests, and how to match users with appropriate products and services.
In this topic, Ma Jianqiang, a senior researcher at Tencent and the person in charge of the online video knowledge graph, will give a keynote speech on vertical search with Tencent video search as the background, including: The main aspects of video search Technology scenarios, algorithm architecture and progress, short video vector recall, application of long video IP knowledge graph, end-to-end search and other cutting-edge technology trends.
With the sudden rise of short video traffic, merchants have seen business opportunities in external media traffic, but direct investment has problems such as high cost of post-link effect analysis. In order to better serve merchants, Alimama's advertising algorithm team has implemented machine learning methods into large-scale data application scenarios of Alibaba's search advertising platform to improve the effectiveness and efficiency of the system. The team's open source Euler graph deep neural network framework and hyperbolic space deep neural network framework are now used by a large number of industry partners and researchers. In this topic, Wang Liang, a senior technical expert in Alimama’s Advertising Product Technology Department and head of external advertising technology, will give you an in-depth explanation of the application of federated learning in Alibaba’s advertising, and a layer-by-layer analysis of the architectural route of Alibaba’s federated learning framework EFLS.
Recommendation system gradually It has become the main way to help people filter information and discover interests. Sequence recommendation aims to use the user's historical behavior to predict the next interaction, but there are implicit and noisy preference signals in the user's long-term historical behavior, which will reduce the modeling effect of the user's true interest. To solve this challenge, Kuaishou and Tsinghua University proposed a sequence recommendation framework SURGE based on graph neural networks. This model provides a new perspective for dealing with sequence recommendation problems and has also achieved huge gains online. In addition, Kuaishou and Renmin University of China proposed a model-independent causal learning framework IV4Rec, thereby enhancing the effect of the recommendation model.
In this sharing, Zang Xiaoxue, a senior recommendation algorithm expert at Kuaishou, will bring Kuaishou’s latest research on causal inference and graph neural network algorithms. These studies have been published at top international academic conferences, and related algorithms have also been published. Implemented in Kuaishou's actual recommendation scenarios, significant online business benefits have been achieved.
ClickAISummit Global Artificial Intelligence Technology Conference Official WebsiteOr scan the QR code below, follow the prompts to completely fill in and submit the information to complete the registration.
The above is the detailed content of How powerful is AI-driven search and recommendation?. For more information, please follow other related articles on the PHP Chinese website!
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
