So I read a lot of literature and technical science popularization, and also consulted the technical students who are surrounded by AI, from which I got a general understanding of some principles of "knowledge map" and compiled the following articles.
I hope my article can help non-technical product managers, or students in other positions, understand what "knowledge map" is more simply and quickly.
Before introducing knowledge map, let's talk about the use of knowledge map in daily life.
For another example, in the online medical industry, when patients want to register but don't know which department to hang, they can get the department information through the pre-diagnosis assistant. Based on the professional medical knowledge map, the pre-diagnosis assistant uses a variety of algorithm models and rounds of intelligent communication to understand the patient's condition and accurately match the medical departments according to the patient's condition.
Take Alipay as an example. In the payment scenario, using the knowledge map to kill bill fraud, credit card cashing and other behaviors in the cradle. Through the map database of knowledge map, the association analysis can be made for different individuals and groups, and users can be judged from the behavior of characters within a specified time, such as IP addresses of places they have been to and used MAC addresses (including mobile phones, PCs, WIFI, etc.). ), correlation analysis of social networks, and whether there is historical transaction information between bank accounts.
Before describing the definition, let's take a look at the representation of the knowledge map 3354 [E-R diagram]:
From the above figure, it can be found that no matter what shape and appearance the E-R diagram is transformed into, it is a relational network formed by connecting multiple points and lines.
We call it point [entity] and line [relationship], and each entity may be related to one or more entities. Based on this, to form the simplest relationship network, only three elements are needed: two entities and one relationship. This structure is called "triplet", and multiple triplets form a knowledge map.
(three times)
For example, "Xiao Fang and Xiao Ming are colleagues, and both of them need to buy notebooks because of their work." Xiao Ming thought it would be more convincing to use Apple notebook, so he started, while Xiao Fang thought Lenovo notebook was cheaper, so he chose Lenovo. Later, Xiao Fang found that the software sketches that his colleague Amway had seen were only available on Apple. It is smarter and easier to use than Axure. "From this sentence, we can disassemble multiple triples:
Knowledge map triples can not only express the relationship between entities, but also express some attributes of entities. For example, "Xiao Ming" is an entity, and his "gender, date of birth and place of origin" can be classified as attributes.
Things are defined as "attributes" of entities, and there are two basic principles:
At the same time, it is worth noting that, according to the actual situation, entities can sometimes be attributes, and attributes can also be entities.
The following figure is an example: "employee" is an entity, and "employee number, name and age" are the attributes of employees. If a "job title" is not linked to "salary, post allowance and welfare", in other words, it has no features that can be further described, then it can be regarded as an attribute of an employee entity according to the standard of 1.
However, if different titles have different salaries, post allowances and different fringe benefits, it is more appropriate to treat titles as an entity.
Having said that, you should be able to better understand the definition of knowledge map: knowledge map is a structured semantic knowledge base, which is used to describe concepts and their relationships in the physical world in the form of symbols. Its basic constituent units are "entity-relationship-entity" triplet, as well as entities and their related attribute-value pairs. Entities are connected with each other through relationships, forming a network knowledge structure.
Understanding the construction of knowledge map can help us better understand the use principle of knowledge map.
The construction process of knowledge map can be summarized in three ways:
In order to introduce each step and its significance, I have compiled the following table:
Please indicate the source for non-commercial reprints.
The following figure is the technical framework of knowledge map, which can help you better understand the process of knowledge map construction. The part in the dotted box is the process of knowledge map construction and knowledge map update.
1) What kind of data is needed to build a knowledge map?
The answer is: structured data.
Generally speaking, there are three kinds of original data of knowledge map: structured data and unstructured data.
The so-called structured data refers to highly organized and neatly formatted data, which is a data type that can be put into spreadsheets. Typical structured data include: credit card number, date, financial amount, telephone number, address, product name, etc.
In contrast, unstructured data refers to data that is not easy to organize or format. It has no predefined data model, so it is not convenient to represent data with two-dimensional logical tables of the database. It can be textual or non-textual, artificial or machine-generated.
Simply put, unstructured data is data with variable fields, mainly some documents, documents and so on. For example, some contract documents, articles, PDF documents, etc.
Semi-structured data is non-relational and has basic fixed structure patterns, such as log files, XML documents, JSON documents and so on.
For unstructured data and semi-structured data, we need to confirm what information can be extracted from them and formulate information entry rules. With the help of NLP and other technologies, effective information can be generated into structured data, and then the structured data can be incorporated into the knowledge map.
2) The difference between graphic database and relational database
Knowledge map is based on graphic database to store data. The so-called graphic database refers not to the database that stores pictures and images, but to the database that stores the data structure of graphics. The E-R diagram we talked about before is the visual display of graphic data. about
Unlike traditional relational databases which use two-dimensional tables to store data, graph databases are traditionally classified as NoSQ.
L (not just SQL) database, that is to say, graphic database belongs to non-relational database. In order to avoid being too technical, I will not introduce the graph data in depth here, but simply talk about the differences between the following databases and relational databases.
Relational database is not good at dealing with the relationship between data, while graph database is flexible and efficient in dealing with the relationship between data.
Traditional relational databases have poor performance when dealing with complex relational data, because relational databases realize relational references between multiple tables through foreign key constraints. Querying the relationship between entities requires JOIN operation, which is usually very time-consuming.
The original design motivation of graphic database is to better describe the relationship between entities. The biggest difference between graph database and relational database is that there is no index adjacency. Each node in the graph data model will maintain its neighbor relationship, which means that the query time has nothing to do with the overall size of the graph, but only with the number of neighboring points of each node, which makes the graph database maintain good performance when dealing with a large number of complex relationships.
In addition, the structure of the graph determines that it is easy to expand. We don't have to consider all the details at the beginning of the model design, because it is easy to add new nodes, new relationships, new attributes and even new labels in the future, and it will not destroy the existing query and usage functions.
In the relational database, if the data fields are designed at the beginning, it will be very troublesome to add more fields after running for a period of time. Developers or product managers need to imagine the fields that may be added in the future at the early stage of development and add them to the data table in advance.
Secondary graphics database
An easy-to-understand knowledge map.
What is a graphic database?
The title map comes from Unsplash and is based on CC0 protocol.
Related question and answer: PC side, what does it mean? PC terminal is a noun corresponding to mobile terminal, which refers to a port that can be connected to a computer host in the network world. It is a computer-based interface system, which is different from the mobile phone interface system of the mobile terminal. In fact, the full English name of PC is: Personal Computer, which translates into Chinese as: personal computer or personal computer. PC is a word with a wide meaning, and it is also a general term for computers. At present, there are many kinds of personal computers, such as traditional desktop computers, DIY computers and notebook computers, as well as tablet computers, all-in-one computers, ultrabooks, palmtop computers and embedded computers that have become popular in recent years. In other words, PC is a broad word, which belongs to the floorboard of computers.