A semantic network, also known as a frame network, is a knowledge base that depicts the semantic relationships between concepts in a network. It is a standard method of knowledge representation. 

Semantic networks gained popularity in artificial intelligence and natural language processing because they represent knowledge or facilitate reasoning. These serve as an alternate kind of knowledge representation to predicate logic.

It is a directed or undirected network composed of vertices, which represent concepts, and edges, which reflect semantic interactions between concepts and map or connect semantic fields. For example, a semantic network may be realized as a graph database or idea map. Semantic triples are used to express typical standardized semantic networks.

Semantic networks are utilized in natural language processing applications, such as semantic parsing and word-sense disambiguation. However, we can also use it to analyze massive texts and identify the key themes and topics (e.g., of social media posts), to expose biases (e.g., in news coverage), and to map an entire research field.

Overview

Richard H. Richens of the Cambridge Language Research Unit implemented "Semantic Nets" for propositional calculus on computers for the first time in 1956. It is an "interlingua" for machine translation of natural languages. However, the significance of this work and the CLRU was not recognized until much later.

Structure

Semantic networks are graphical representations of information composed of nodes and links that illustrate the hierarchical relationships between items. The nodes are connected via links or arcs. These arcs represent the connections between the various items and descriptors.

A semantic network organizes information better understood as a group of connected ideas. Most semantic networks are built on the way people think. For example, the ideas of activation spreading, inheritance, and nodes as proto-objects came from semantic networks.

One way to make semantic networks, sometimes called co-occurrence networks, is to detect keywords in the text, figure out how often they appear together, and then look at the networks to find key terms and groups of related themes.

Usage

A semantic network is used when one possesses knowledge best understood as a group of related concepts. The majority of semantic networks are cognitive. They also have arcs and nodes that can be arranged in a taxonomic hierarchy.

Benefits

Humans find semantic networks easier to read and comprehend. They are more efficient and easier to deploy. (since they can use special purpose procedures). In specific ways, they can be more expressive than first-order logic. (for instance, inheritance with exceptions).