Difficulties in the development of knowledge map & the importance of building industry knowledge map

One 、 summary

Although AI has made rapid progress relying on machine learning and deep learning , But these are weak artificial intelligence , For machine training , It needs human supervision and a lot of data to feed , What's more, people need to mark the data manually , For strong AI , This is not desirable . To realize real human like intelligence , Machines need to master a lot of common sense knowledge , Language understanding is based on people's thinking mode and knowledge structure 、 Visual scene analysis and decision analysis .

Two 、 What is knowledge map

Baidu Encyclopedia definition ： Knowledge map is also called scientific knowledge map , In the field of Library and information, it is called knowledge domain visualization , Or knowledge domain mapping map , A series of different graphs used to show the relationship between the process of knowledge development and structure , Using visualization technology to describe knowledge resources and carriers , mining 、 analysis 、 structure 、 Draw and show knowledge and their relationship to each other .

Knowledge map of breast cancer

3、 ... and 、 Why need knowledge map

1、 Knowledge atlas makes complex knowledge domain and knowledge system through data mining 、 Information processing 、 Knowledge measurement and graphic drawing are displayed , Indicates the development trends and laws in this field , Provide all-round information for the research in this field 、 entirety 、 Reference to the relationship chain .

2、 Knowledge map is an important means of production in intelligent society , If artificial intelligence is compared to a “ The brain ”, So deep learning is “ The brain ” How it works , The knowledge map is “ The brain ” The knowledge base of , And big data 、GPU Supporting technologies such as parallel computing and high-performance computing are “ The brain ” The support of thinking operation .

3、 Knowledge map is the semantic representation of the real world , Each node represents the edge of the entity connection node, which corresponds to the relationship between entities. Heterogeneous data is expressed as knowledge through integration , The expression of the graph reflects the human cognitive way of the world , Knowledge map is very suitable for integrating unstructured data and discovering knowledge from scattered data , So as to help organizations realize business intelligence .

Four 、 The birth of the knowledge map

When you study in an unfamiliar field , Unable to grasp the key points and the overall framework , So that the retrieval efficiency is low 、 When there is no Tao at the beginning , The knowledge map came into being .

since 2012 year 5 month ,Google Apply knowledge map to its search engine , To improve its search service capability , Display the relevant information collected from various channels in the information box next to the search results , It is provided to users in the form of structured modules .

It improves its search effect mainly from three aspects ：

1、 The polysemy of language , Show differentiation results , Narrow your search .

2、 Relevance of information , Understand the relationship between summary information , Understand the relevance of things .

3、 The universality of the system , Build a complete knowledge system , Discover new facts or new connections , Promote a whole new set of search queries .

google Knowledge map display

5、 ... and 、 The principle of knowledge map

Knowledge maps are processed with structured data , Use the ternary , spot 、 Line 、 Faces represent the relationship between ontologies , Use relationships to organize all objects ( Entity ), Form a directed graph structure . Knowledge , It refers to the information corresponding to the point or edge .

Knowledge map is based on semantic analysis technology , Model centered , Data based , Using deep neural networks 、NLP Intelligent processing technologies such as framework semantic understanding are applied to input words 、 word 、 Chapters are carried out at multiple levels 、 Multidimensional information analysis , Provide remotely callable entity extraction 、 Algorithm service interface capabilities such as relationship extraction and attribute extraction . To build a multi domain knowledge map platform , Serve different industries and application scenarios .

Knowledge map related technology

Building a knowledge map is an iterative process , According to the logic of knowledge acquisition , Each iteration can be divided into three stages ：

1、 Information extraction ： Extract entities from various types of data sources 、 Properties and relationships between entities , On this basis, ontology knowledge expression is formed ;

2、 Knowledge fusion ： After acquiring new knowledge , It needs to be integrated , To eliminate contradictions and ambiguities , For example, some entities may have multiple expressions , A certain appellation may correspond to many different entities, etc ;

3、 Knowledge processing ： For new knowledge that has been integrated , After quality assessment （ Some of them need to be screened manually ）, In order to add the qualified part to the knowledge base , To ensure the quality of the knowledge base .

Knowledge Mapping Technology Architecture

6、 ... and 、 The development direction of knowledge map

“ Pure universal AI doesn't make any sense , The future direction of artificial intelligence must be industrialization .”

Therefore, the construction of industry knowledge map is particularly important , First , The construction of industry knowledge atlas system should be based on the aggregation and fusion of massive data 、 Rapid perception and cognition 、 Powerful analysis and reasoning 、 Self adaptation and self optimization and industry intelligent decision-making .

So let's talk about that   General knowledge map   And   Industry knowledge map   Differences in construction ：

The universal knowledge map is based on Internet open data , For example, Wikipedia or community crowdsourcing is the main source , Gradually expand the scale . Based on triplet factual knowledge , More open domain oriented Web extract , There is a certain tolerance for the quality of knowledge extraction , Improve data quality with knowledge fusion , The application fields are mainly in search and question and answer , Lower requirements for reasoning . Such as ： Baidu 、 Google 、 Yahoo, etc

The industry knowledge map takes the data within the field or enterprise as the main source , It usually requires rapid scale-up , Build industry barriers , The knowledge structure is more complex , It usually includes ontology engineering and regular knowledge . The quality of knowledge extraction is very high , It depends more on the internal structure of the enterprise 、 Joint extraction of unstructured and semi-structured data requires manual verification , To guarantee the quality . Areas that usually need to integrate multiple sources are effective means to expand the scale of data . The application form is more comprehensive , In addition to search questions and answers , It also includes decision analysis 、 Business management, etc , And higher requirements for reasoning , And has strong interpretability requirements . The main fields are e-commerce 、 Finance 、 Agriculture 、 Security 、 Medical treatment, etc.

Construction of industry oriented knowledge spectrum system

The collection, analysis and computing power of big data is no longer an obstacle , The difficulty lies in the construction of knowledge map behind the algorithm . In the actual , Flexible enterprise data and business changes , So the data source 、 data structure 、 Data content changes at any time , The understanding of business and the interpretation of data also change . therefore , Build real-time agility 、 Flexible and scalable 、 Intelligent adaptive dynamic knowledge map is particularly important .

7、 ... and 、 The importance of the industry knowledge map

because IT The rapid development of the times , It forms the aggregation of data . Promoted the coming DT Time , Data and algorithm provide new possibilities for the construction of knowledge map , And the knowledge map as AI Support foundation , Although the development is slow , It is the only way for AI . Because of its semantic search 、 Intelligent q&a 、 Data analysis 、 natural language processing 、 Visual understanding 、 Internet of things devices have shown increasing value .

such as ： The robot based on voice dialog system mentioned in the previous article , If you want robots to be fun , It is necessary to establish a general knowledge map , Make the robot have powerful logic , Understand the relationship between people , The relationship between people and things , Human attributes 、 The nature of things , Understand the chain of relationships in the human world , Flexible response to various chat needs of users , Make the machine have humanoid intelligence . The industry knowledge map helps users answer the task-based needs of the industry , Assisted user decision making , Reverse helps people learn .

Now , Artificial intelligence belongs to the early stage of development , The development of the industry is mainly driven by technology , It belongs to the period of looking for nails with a hammer , Enterprises in the industry +AI As the core , We should also pay attention to the industry + Knowledge map . Because the future technology is certainly not the core competitiveness of the company , The industry data accumulated over the years is the barrier . Data nurturing AI,AI Feedback data .

The industry knowledge map has been well applied in many fields .

for example ： Check the inner eye 、 Enterprise knowledge map of enterprise search , The data includes ： Enterprise basic data 、 Investment relations 、 Employment relationship 、 Enterprise patent data 、 Enterprise bidding data 、 Enterprise recruitment data 、 Enterprise litigation data 、 Enterprise dishonesty data 、 Enterprise news data ;

Enterprise knowledge map

Use the knowledge map to integrate the above data , Making enterprise knowledge map , And make use of the characteristics of the enterprise knowledge map , Make a series of applications for financial business scenarios .

1、 Enterprise risk assessment

Based on enterprise basic information 、 Investment relations 、 litigation 、 Dishonesty and other multidimensional related data , Using graph calculation and other methods to build a scientific 、 Rigorous enterprise risk assessment system , Effectively avoid potential business risks and capital risks .

2、 Corporate social graph

Based on investment 、 In office 、 patent 、 Tender unit 、 The litigation relationship takes the target enterprise as the core and spreads to the outside , Form a network diagram , Intuitively and stereoscopically display enterprise associations .

3、 The person in charge of the enterprise

Looking for the shareholder with the largest shareholding ratio based on the equity investment relationship , Finally, it can be traced back to natural persons or state-owned assets management departments .

4、 Enterprise Association path

Based on equity 、 In office 、 patent 、 Tender unit 、 In the network relationship formed by the litigation relationship 、 Query the shortest path between enterprises , Measure the closeness of the relationship between enterprises .

5、 The development course of the enterprise

Based on the time sequence of investment and financing events in the enterprise knowledge map , Record the development history of the enterprise .

6、 Intelligent question and answer of enterprise information

Users input through voice , The system outputs the desired answer to the user through voice .

That's all , I hope I can give you some thoughts . About the medical knowledge map 、 Financial knowledge map 、 Please check the application scenarios of agricultural knowledge map by yourself .