(1) Descriptive knowledge: levels of classification and attribute inheritance.
Describe the attributes and relationships of things, things, things/things, that is, descriptive knowledge (that is, conceptual knowledge, proposition knowledge, or statement knowledge). Expand its original definition by deep descriptive knowledge, if used (appropriate) or conceptual layers. This knowledge can include facts and recording systems. It can be organized, used and updated the facts and information related to the environment.
The basic subject (Ontology) used in a single artificial intelligence system can be sowed from the type and entity related to task -related types (e.g., an entity type named by Opencyc or AMR). This basic subject can be expanded through neural network/machine learning technology -that is, if you get new knowledge, you will get new entities, relationships, and categories.
(2) World model.
The artificial intelligence system can understand the situation, explain the input/event, predict the potential future results, and take action. The phenomenon model is an abstract/general, which can be divided into an approximation (informal) model of the formal model and the real world; the phenomenon model allows the use of variables and applications in specific conditions, and allows symbols to operate specific instances or more commonly common cases or more commonly operates category.
Examples of formal models include logic, mathematics/algebra, and physics. Models in the real world are usually experienced, experiments, and sometimes even chaotic, not formal. The models in the real world include physical models, psychological models and sociological models. This category includes program models (proprietary knowledge).
Cause and effect models can help artificial intelligence systems to develop a higher level. If the context changes, if the past statistical data is combined with the knowledge model, if the context of causality, understanding and control causes, and the ability to consider the opposite facts can be effectively applied to predict the future. From the perspective of conditions and possibilities, these models help understand the situation or incident. In this way, machine intelligence can reach the level of human intelligence.
(3) Stories and scripts.
As historian Yvar Hallari said, this story is a key component of individuals, society, culture, and world views. The concept of the story must fully understand and explain human behavior and exchanges. This story is very complicated, and multiple events and various information may be included in a coherent narrative. This story is not only a collection of facts and events, but also contains important information other than developing understanding and summarizing data. Unlike the world model, this story can be regarded as historic, reference, or spiritual significance. This story can represent values and experience, which can affect people's beliefs and behaviors. These examples include religious or ethnic stories, myths, and stories shared at any level.
(4) The context and attribution of the source.
The definition of context is a framework, the framework of the event, and provides resources itself. The context can be regarded as a knowledge structure that covers the knowledge it contains. The context can be lasting or short.
According to the new learning materials, the lasting context can be long -term (such as knowledge obtained from Western philosophy or Eastern philosophy), or it can change over time. The lasting context will not change each task.
When a specific local context is very important, the transient context is related. These words are explained according to the local context of the surrounding sentences or paragraphs. In the context of the entire image or video, the areas of interest are usually explained in the image.
The combination of lasting context and transient context can provide complete interpretation and operation knowledge settings.
Another related aspect of knowledge is data source (also known as data traceability), which includes data source, what happened during the data transmission, and where the data will go over time. The artificial intelligence system cannot consider all the information to be correct or credible, especially when it is called the post -truth era. Establishing credibility, certification and traceability may need to connect information with its source.
(5) Value and priority (including good/threatening and ethics)
Throughout the scope of judgment, all aspects of knowledge (e.g., objects, concepts, or procedures) have corresponding value -from the maximum good to the greatest evil. It can be considered that the evolution of human intelligence includes the pursuit of returns and avoiding risks (for example, pursuing lunch; avoiding it as lunch). This risk/return is closely related to knowledge. Potential benefits and losses have utilitarian value; it also has moral value based on entity or potential future state. This moral value reflects a moral value, that is, good is not based on potential tangible returns or threats, but based on what is the right potential belief lte iot.
The subjective judgment of the artificial intelligence system on knowledge, actions and results reflects the value and priority of the artificial intelligence system. This laid the foundation for the accountability system and should be handled by people responsible for a specific artificial intelligence system. When an artificial intelligence system interacts with humans and selects human benefits, the potential value and priority system is very important.
ConceptReFerferrence (CONCEPTREFEFE) is a symbol and reference set for all things related to given concepts. The concept quotes does not actually contain any knowledge -knowledge stays in the dimension introduced earlier. Conceptual reference is the key to the multi -dimensional knowledge base (KB), because concept references integrate all the appearance of the concept.
Wikidata is a good example of centralized storage structured data for multi -dimensional knowledge bases. In Wikidata, items (items) represent everything in human knowledge, including themes, concepts, and objects. Wikidata's project is similar to the definition of CONCEPTREF in this framework -there is only one key difference: in Wikidata, the term items refer to not only the given symbols, but also the information related to symbols; ConceptRefs is just a sign symbol, pointing to the KB pointer. The various views described in the previous chapter fill the information about concepts (such as descriptive or programmatic knowledge and concepts).