PhD student at Carnegie Mellon University Ross Qullian showed semantic networks could use graphs to model the structure and storage of human knowledge. Quillian hoped to explore the meaning of English words through their relationships. The graphs featured nodes connected by lines. Each node represents a concept or word the network “knows,” and a line between nodes represents a relationship. They included class (relationships by “is” and “a” such as “A canary is a bird.”), modification (adjectives or adverbs), conjunction (“and”), and disjunction (“or”). Quillian and cognitive scientist Allan Collins would later add proximity (distance between nodes), consequence (what a node causes), precedence (priority between relationships), and similarity. They described this in their 1972 paper “How to Make a Language User” and Quillian’s 1968 paper “Semantic Memory.”
Qullian developed a method for searching the semantic network. First, he would compare two nodes and create mappings to each other node. Then, he would store each mapped node and describe their relationships to the compared node. Creating labels for each relationship, they did this until all relationships were formed.
Later scientific research showed this model simulates reaction time for how humans compare relationships in proving statements. Humans and Quillian’s model both interpret “Canaries have feathers” faster than “Canaries have skin.” The diagram of the relationships shows this:
“has Feathers” is a relationship of “Bird,” while “has Skin” is from “Animal.” When reasoning, humans and computers both arrive at “Feathers” first because “Bird” is closer to “Canary” than “Animal” is.