the era of "atlas"

Knowledge atlas has been fermented since 212, and it has become increasingly fierce. The top leaders in the industry have released the application of enterprise knowledge atlas, which can realize data value for enterprises. It can only be said that with the rapid development of graphic technology, knowledge map is an irresistible trend whether the business needs change or not. On April 2, 22, the National Development and Reform Commission defined the connotation of "new infrastructure" of artificial intelligence, reflecting the characteristics of "emphasizing innovation and supplementing shortcomings": helping the intelligent transformation of traditional infrastructure and improving the operational efficiency of traditional infrastructure design.

figure 1? Growth scale of benefit of knowledge map in China —— iResearch

At present, artificial intelligence can be simply divided into perceptual intelligence (mainly focusing on exploring the ability of pictures, videos and voices) and cognitive intelligence (involving knowledge reasoning, causal analysis, etc.).

artificial intelligence is the key area of new infrastructure, and knowledge map is the underlying support of cognitive intelligence. Knowledge map has the ability to interpret data, reason and plan a series of human thinking and cognition, and is based on large-scale and highly correlated background knowledge.

? -White Paper on Knowledge Mapping Industry for Artificial Intelligence "New Infrastructure"?

We use knowledge maps every day

Knowledge maps are applied in various fields, such as e-commerce (product recommendation), medical treatment (intelligent diagnosis), finance (risk control) and securities (investment research). Well-known enterprises include: Google Knowledge Graph, Meituan Brain, Alibaba Tibetan Scripture Pavilion Project, Tencent Cloud Knowledge Map TKG, etc.

knowledge map plays an important role in many fields of artificial intelligence: semantic search, intelligent question answering, assisting language understanding, assisting big data analysis, enhancing the interpretability of machine learning, and assisting image classification by combining graph volume product. At the same time, it also means that the technical difficulty is greatly increased.

the value of knowledge map

you may think that knowledge map is the ultimate goal of capturing and managing knowledge. In fact, knowledge graph is good at explicitly capturing knowledge in a top-down relational connection way. Connect upstream and downstream relationships through relationship nodes, and clearly sort out the relationship network. As shown below:

Figure 2? In pervasive intelligent knowledge, the platform

efficiently and intuitively depicts the correlation network between target subjects (such as enterprises and events), thus depicting enterprises in all dimensions and reproducing the real situation and complicated relationships of subjects in three dimensions. Its powerful interconnected organization ability and visual decision-making reasoning support provide the underlying foundation for enterprise assets. One-stop application of pervasive intelligence "graphic intelligence" has the ability to open all kinds of tricks, with the following considerations:

Deep link analysis can be found organically

Take our most familiar financial field as an example, the common entities of knowledge map include companies, products, personnel and related events, and the common relationships include equity relations, employment relations, supplier relations, upstream and downstream relations, competition relations and so on.

the advantage of doing this is that through the integration of knowledge maps, the original complex data will form an intuitive and easy-to-understand visual map. Under the trend of global economic integration, analysts and investment institutions are likely to observe the changes in the competitive landscape one step ahead, providing clues for finding new customers and new investment opportunities.

figure 3? What are the

multi-dimensional attributes of enterprise upstream and downstream relationship networks? Follow the trail

Another value of knowledge map is that it can simply handle multi-dimensional data. At present, Ubiquitous Intelligence helps customers analyze billions of entities (or nodes) and relationships (or edges).

Figure 4: Screenshot of a joint-stock commercial bank's fund product relationship network

"For beneficial ownership, we often see ownership classes with six, seven or more floors, especially in places with large enterprises like China." ? "People must realize that a good tool that can handle and query at least six to seven layers (if there are no more layers) is the real core of solving problems."

every company, individual and news event can be a "point", and the artificial intelligence engine can aggregate these points, analyze the correlation, similarity and aggregation degree from multiple dimensions, and restore the real scene, so as to "follow the trail".

Figure 5 Application of Anti-fraud Map

For example, knowledge map In the traditional risk management process, it is impossible to judge the real associated risk by strictly examining the characteristics of the simple dimension of the target subject.

Challenges and Opportunities

Pervasive intelligence is deeply rooted in the financial field, and its sub-business scenarios include but are not limited to: anti-fraud, anti-money laundering, stolen brush investigation, lost collection, abnormal monitoring of foreign exchange, credit audit, etc. Take a specific project as an example: due to the long process of map construction itself and the relative independence of map construction in each scenario, it creates necessary conditions for repeated data development and data disconnection, but a large number of enterprise assets cannot be bypassed.

figure 6? The construction mode of traditional relational network application

In terms of project landing, there are still some problems, such as long construction cycle of atlas, high professional degree of application construction and high cross-industry migration cost. The challenge will be reflected in whether the product can be used out of the box.

the idea of platformization in pervasive intelligence

in order to solve the above problems, pervasive intelligence independently developed and upgraded the knowledge map construction and application platform to a one-stop "graphic intelligence" middle platform.

figure 7? The construction mode of traditional relational network application

A set of platform of middle platform and factory mode is born, which ensures the demand of various scenarios for different forms of graphs and joint query. "Know everything, know everything", and the one-stop "Graphic Intelligence" platform is "that skill", as follows:

Open up the independent map construction of business scenarios, reduce the repeated development cycle cost, empower traditional application forms, improve service quality and efficiency, simple map applications can be realized within 1-2 days, and complex map applications can be shortened to one-third in traditional practices, accelerating the accumulation of enterprise assets;

cooperate to get through departmental data, and solve the problem of long communication cycle and difficult cooperation between departments;

The atlas is interactive and friendly, and it is easier to find hidden information by making visual decisions to assist business scenarios;

Empower experts and industry experts, and program the industry experience of domain experts on the platform, and enterprise knowledge assets will be precipitated.

Real-time scalability and flexibility

The value of the platform in the knowledge map also lies in its flexibility and extensibility, and the establishment of a real-time agile, flexible, extensible and elastic data base. The financial knowledge map directly feeds back the rigid demand of the financial industry. In practice, enterprise data and business change flexibly, and the data source, data structure and data content will change at any time, so will the understanding of business and the interpretation of data.

figure 8? Multidimensional data expansion query

How to use these data effectively requires employees to have professional financial knowledge and deeply understand the correlation and conduction that may be caused by a certain data change. Knowledge map will be the most handy tool.

Graph technology is the strongest ammunition for the application of knowledge mapping

Enterprises need to be able to quickly support new iterative modes in business. The middle platform of "graph intelligence" with pervasive intelligence has computing engines: graph computing model and graph matching business data model, which help enterprises to achieve this goal.

graph rule calculation: (for example, a customer who uses a phone with a blacklisted customer * * * is a suspected fraudulent customer)

graph index calculation: (for example, the proportion of blacklisted customers within the customer's two-degree relationship)

graph machine learning (in order to make feature engineering more effective as a priori knowledge)

community identification: tag prediction (blacklist prediction/potential VIP customer prediction) < p Community analysis

Shortest path: Optimize the processing path and save the data processing cost.

figure 1? Path query

"If a worker wants to do a good job, he must sharpen his tools first". The one-stop application of pervasive intelligence "graphic intelligence" provides effective methods and tools for describing the production and life behavior of the physical world. Gartner: "The era of drawing has arrived", let's "draw" together!