The explosive growth of information on the internet, characterized by large-scale, diverse and loose organizational structures, poses a huge challenge to effective access to information and knowledge。
The “knowledge mapping” that has emerged in recent years has gradually become an innovative way of managing big information, favoured by major internet companies, and applied unwittingly widely in various fields, such as search, electronics, socialization, etc。
In addition, the technology has been brought to the attention of major financial institutions and artificial intelligence companies and has begun to be applied on a pilot basis in new areas such as financial fraud and security。
However, knowledge mapping techniques are still underdeveloped and commercialization, such as the acquisition of fees for technical services from third parties, is still at an exploratory stage。
I. Basic concepts of knowledge mapping
With regard to the definition of knowledge mapping, professor xiao yihua of the university of jordan explained this in his book knowledge mapping: “as a form of expression of knowledge, knowledge mapping is a large-scale semantic network that includes entities, concepts and the various semantic relationships between them.” as shown below, plato is an entity and he is a philosopher (concept)。
“the core value of knowledge mapping lies in how knowledge is deposited. At the beginning of the creation of the spectra, it was a relationship schematic that was built and formed through a large number of models, drawing on the highly interpretative advantages of the spectrograph model, and eventually becoming a knowledge schematic. In the process of building the relationship map to the knowledge map, operational experts are highly involved, using the exploratory data analysis capacity of the knowledge map, self-help in the construction of highly interpretative business knowledge models and the formation of an iterative and high sharing of knowledge deposits.”
Knowledge mapping is a typical large-data age product with enormous, high-quality, structurally friendly advantages。
In 2017, when the country's scientific catalogue was adjusted, for the first time a knowledge mapping discipline emerged, and the ministry of education positioned the knowledge mapping discipline as “large-scale knowledge engineering”。
Knowledge mapping development and application
The early use of knowledge mapping is mainly for the optimization of search results, and applications are increasingly being explored with the rise of electricians, internet finance, etc., including smart recommendations, smart customer service, financial weather control, safety and security。
1. Search optimization
Knowledge mapping dates back to around 2010, mainly to address the user experience of the search engine. The search area faces two main problems:
On the one hand, search needs and results are often difficult to match, and there are frequent “searches” of non-questions; on the other hand, the results of searches are poorly organized and show confusion。
In 2010, microsoft began building microsoft satori knowledge mapping to enhance bing's search capability; in may 2012, google launched knowledgegraph to support its semantic search and is now the largest knowledge mapping in the world. On 22 november 2012, the search for dogs over the cube became the country's first search engine, the chinese knowledge map. At about the same time, 100-degree knowledge maps were set up. In 2013, facebook released opengraph for social networking intelligence search. Early large search platforms rely mainly on “keyword search” techniques, returning to the list of web pages containing keywords for users, who need further browsing of these pages and filtering out a large amount of useless information to find what they really want, and who would prefer “direct answers”. Knowledge mapping techniques can directly give the search results that users want, rather than links. Here's the figure: "how many people in shanghai?" the 360 search shows the data directly from the national statistical office, and the user moves the mouse directly to the corresponding year and can quickly view the data for each year。
Knowledge maps were then introduced into the electrical search field. In 2015, alibaba began to construct a knowledge map for the electrician sector — the cognitive map — and in 2016 amazon also followed closely with the construction of the knowledge map. In may 2018, the nlp center began to construct a large-scale knowledge map of catering and entertainment - the mission brain。
In the search area, the knowledge mapping enables users to understand the query terms submitted as an entity or a concept and to return to the full range of web content that the user may be interested in by means of physical or conceptual matching. The search process for knowledge maps in the search engine is as follows:
Knowledge mapping uses text and knowledge-integrated reading understanding models to make search results more understandable to clients. For electrical platforms, the volume of transactions and the level of customer activity are also core competitiveness, while customers generally obtain the desired goods by searching, and the more accurate the search results, the more they use them. As a result, hundreds of searches, dog searches, ali baba, the united states of america, and so on have attempted to build their own knowledge mapping platforms。
2. Smart recommendations
In addition to optimizing search results, knowledge mapping can also help electricians and social platforms to address some of the smart recommendations. For example, some of the current small and medium platforms have the biggest problem with smart recommendations: “what to buy, what to recommend” or “what to recommend is not related to customers”. The lack of novelty in the recommended commodities results in a general transformational effect. Knowledge mapping can help the electrician platform to escape this simple recommendatory logic, making the recommended results more intelligent and facilitating user purchase。
Comment: the left chart shows how entity links can be applied to smart question and answer recommendations, which are addressed through slus with the intention and principal entity, and then with the entity's recommendation, the entity is used to build related issues, for example, “wuhan university is good around?” to identify its main entity as “wuhan university” and its core intention is to be food, and to recommend entity-related issues such as “what is good around china university of technology?” “what is good around qinghua university?” “what is good around wuhan university of technology? And the question of what is going on around wuhan university, for example, is what's going on? "wait。
The right-hand map shows the applications recommended by the entity for the automated mapping platform, for which the relevant entity is recommended when the user searches for an entity。
3. Financial controls
In recent years, following the emergence of consumer finance and micro-enterprise lending, banks and other licensed financial companies, lending institutions, artificial intelligence companies, etc., have begun to apply knowledge mapping to risk control, especially to identify group fraud。
The reasoning and interpretability of knowledge maps have a natural advantage in the financial landscape. The 2020 china knowledge mapping industry study for artificial intelligence “new capital” released at the beginning of the year by irei states that, with the explosive growth of financial data in recent years, traditional wind control systems have gradually failed, while smart wind control systems that apply machine learning algorithms and knowledge mapping have a prominent advantage in risk identification capabilities and large-scale computing, becoming the main means of controlling fraud in the financial sector。
In the area of finance, knowledge mapping can be applied to anti-fraud operations such as micro-enterprise credit, consumer credit and credit card applications, and can also be used to identify accounting fraud。
The rationale is simple to understand: “the things are clustered in groups and the humans in groups”. It may be fraud-related, for example, “a multi-business loan client under the same wifi”, or “an application for a multi-business account with the same equipment registered for borrowing”. Thus, the identification of credit fraud can be translated into customer knowledge mapping or social networking analysis. That is, the integration of business business information, news dynamics, shareholder relationships, equity changes, judicial proceedings, etc., into the anti-fraud knowledge map, analysed and predicted, and the identification of fraud cases, such as the use of shell company loans. According to the director of the ice science and technology knowledge mapping, ice mapping is currently working with banks to model knowledge mapping under different conditions and scenarios, labeling the physical relationship of the knowledge mapping or digging new relationships for better application to financial wind control。
Figure 7: design of the overall framework of the graph features
