The Knowledge Graph is an important branch of artificial intelligence. It was proposed by Google in 2012 and has become a killer application for building large-scale knowledge, playing a crucial role in areas such as search, natural language processing, intelligent assistants, and e-commerce. The Knowledge Graph, along with big data and deep learning, has become one of the core driving forces behind the development of the internet and artificial intelligence.
Concept and Classification of Knowledge Graphs
The Knowledge Graph was proposed by Google in 2012 and successfully applied in search engines. It describes concepts, entities, and their relationships in the objective world in a structured form, representing information on the internet in a way that is closer to human cognitive understanding, thereby providing a better capability to organize, manage, and understand the vast amounts of information on the internet.
There are many ways to classify Knowledge Graphs, such as by type of knowledge or construction method. Generally, Knowledge Graphs can be divided into two categories: general Knowledge Graphs and specific domain Knowledge Graphs.
▲Diagram of Knowledge Graph
Common diagrams of Knowledge Graphs mainly consist of three types of nodes: entities, concepts, and attributes.
An entity refers to a distinguishable and independently existing thing, such as a person, a city, a plant, or a product. The world is composed of specific entities, which are the basic elements of a Knowledge Graph, and different entities have different relationships.
A concept refers to a collection of entities that share similar characteristics, such as countries, ethnicities, books, and computers.
Attributes are used to distinguish the characteristics of concepts, with different concepts having different attributes. Different types of attribute values correspond to different types of attribute edges. If the attribute value corresponds to a concept or entity, it describes the relationship between two entities, known as an object property; if the attribute value is a specific numerical value, it is called a data property.
Three Typical Applications of Knowledge Graphs
Currently, internet giants led by commercial search engine companies have recognized the strategic significance of Knowledge Graphs and have invested heavily in their development, which has increasingly influenced the form of search engines. Designing and implementing Knowledge Graph applications based on business needs and optimizing them according to data characteristics is a key research area.
The typical applications of Knowledge Graphs include semantic search, intelligent question answering, and visual decision support.
1. Semantic Search
Current keyword-based search technologies, supported by Knowledge Graphs, can elevate to entity and relationship-based retrieval, known as semantic search.
Semantic search can accurately capture user search intentions using Knowledge Graphs, solving the problems of keyword semantic diversity and semantic disambiguation encountered in traditional search, and achieving mixed retrieval of knowledge and documents through entity linking.
Semantic retrieval needs to address the diversity of expressions brought by natural language input and the ambiguity of entities in language. With the aid of Knowledge Graphs, semantic retrieval needs to directly provide answers that meet user search intentions, rather than links to relevant web pages containing keywords.
2. Intelligent Question Answering
A Question Answering (QA) system is an advanced form of information service that allows computers to automatically answer questions posed by users. Unlike existing search engines, QA systems return precise answers in natural language rather than a ranked list of relevant documents based on keyword matching.
Intelligent question answering systems are considered one of the disruptive technologies of future information services and are regarded as a primary means of validating machine language understanding capabilities.
Intelligent question answering needs to understand the user’s natural language input and provide answers to user questions from Knowledge Graphs or target data, with key technologies and challenges including accurate semantic parsing, correctly understanding the user’s true intentions, and scoring returned answers to determine priority order.
3. Visual Decision Support
Visual decision support refers to providing a unified graphical interface that combines visualization, reasoning, retrieval, etc., to provide users with an entry point for information acquisition. For example, decision support can interpret information such as the development status of start-ups and investment preferences of investment institutions in venture capital graphs through visualization techniques, displaying comprehensive information about companies through node exploration, path discovery, and association exploration.
Key issues to consider in visual decision support include assisting users in quickly discovering business patterns through visualization, enhancing the interactivity of visual components, and improving the efficiency of underlying algorithms in large-scale graph environments.
Five Development Stages of Knowledge Engineering
The Knowledge Graph technology is part of knowledge engineering. In 1994, Turing Award winner and founder of knowledge engineering, Feigenbaum, defined knowledge engineering as integrating knowledge into computer systems to accomplish complex tasks that only specific domain experts can perform.
Looking back at the development of knowledge engineering over the past forty years, we can divide it into five landmark stages: the pre-knowledge engineering period, the expert system period, the Web 1.0 period, the collective intelligence period, and the Knowledge Graph period, as shown in the figure below.
1) 1950-1970: Turing Test—Early Birth of Knowledge Engineering
This stage mainly had two approaches: symbolism and connectionism. Symbolism holds that physical symbol systems are necessary and sufficient conditions for intelligent behavior, while connectionism believes that the brain (neurons and their connection mechanisms) is the foundation of all intelligent activities.
The knowledge representation methods during this period mainly included logical knowledge representation, production rules, and semantic networks.
2) 1970-1990: Expert Systems—Prosperous Development of Knowledge Engineering
Due to the emphasis on utilizing human problem-solving abilities to establish intelligent systems while neglecting the support of knowledge for intelligence, artificial intelligence struggled to play a role in practical applications. Starting in the 1970s, artificial intelligence began to shift towards building knowledge-based systems, achieving machine intelligence through “knowledge base + inference engine”.
During this period, knowledge representation methods underwent new evolutions, including frameworks and scripts, and many expert system development platforms emerged in the late 1980s, helping to transform domain knowledge from experts into knowledge that computers can process.
Between 1990 and 2000, many manually constructed large-scale knowledge bases emerged, including the widely used English WordNet, the Cyc common sense knowledge base using first-order predicate logic, and the Chinese HowNet.
The emergence of Web 1.0 provided an open platform for people, using HTML to define the content of text and connecting texts through hyperlinks, allowing the public to share information. The extensible markup language XML proposed by W3C enabled the structural marking of internet document content through tag definitions, laying the foundation for large-scale knowledge representation and sharing in the internet environment.
4) 2000-2006: Collective Intelligence
The emergence of the World Wide Web shifted knowledge from closed to open, from centralized construction to distributed collective intelligence knowledge. Originally, expert systems contained knowledge defined within the system; now knowledge sources can interlink, generating more knowledge through associations rather than being entirely produced by fixed individuals.
During this process, collective intelligence emerged, with Wikipedia being the most typical representative, where users establish knowledge, reflecting the contribution of internet users to knowledge, becoming an important foundation for today’s large-scale structured Knowledge Graphs.
5) 2006-Present: Knowledge Graph—New Development Period of Knowledge Engineering
“Knowledge is power”; the goal of this period is to transform the content of the World Wide Web into machine-understandable and computable knowledge that can drive intelligent applications. Since 2006, the emergence of large-scale rich structured knowledge resources like Wikipedia and advancements in network-scale information extraction methods have led to significant progress in large-scale knowledge acquisition methods.
Currently, automatically constructed knowledge bases have become powerful assets for semantic search, big data analysis, intelligent recommendations, and data integration, widely used in large industries and domains. A typical example is Google’s acquisition of Freebase and the launch of Knowledge Graph in 2012, along with Facebook’s graph search, Microsoft Satori, and specific knowledge bases in business, finance, and life sciences.
Source: AI Scientist
Recommended Reading:
[November Release – Knowledge Graph] The “New Coronavirus Pneumonia” Diagnosis Graph Debuts and the “Drug Indications” Graph is Updated
Knowledge Graph-Based Q&A
Looking Ahead: The 2021 Core Resource Release Plan for OMAHA has been Announced
The 2021 OMAHA Sixth Annual Conference “Intelligent-Driven Deep Medicine” will be held in January
The first offline training of 2021—Ontology Medical Terminology System and Application Special Training will start on January 15!
OMAHA HiTA: Metadata | Terms | Knowledge Graph
To heal, we choose openness and sharing
Click “Read Original” to Join Us Now, and start the movement to return personal health data!
