Finding words through spreading activation
The associative network of 150 nouns resulting from our adaptive hypertext experiment can be used to implement "spreading activation" (see a technical report on spreading activation models in information retrieval). This means that selected words in the network get an activation value, and that this activation follows the links to associated words. The activation reaching a word is the sum of the activations coming through all its input links, each proportional to the strength of the link. From an activated word, activation diffuses further to the words linked to it, and so on.
This makes it possible to find words which are related to all the words that are initially activated. For example, the activation diffusing from the initial selections "paper" and "education" will concentrate most strongly in the word "book". Other examples are "control" and "society", which activate "government", "car" and "town", which activate "road", and "work", "room" and "building" which together activate "office". This is similar to the way thoughts diffuse in the brain, moving along intuitive, fuzzy pathways, rather than retrieving exact matches like traditional web search engines.
You can try out spreading activation on our results by selecting a combination of words in the following form. If you then click on the "Activate" button, the program will reply with a list of the words that got the highest activation, and a representation of the strength of activation.
Copyright© 1998 Principia Cybernetica -
Referencing this page
|
|
Author
J. Bollen, & F. Heylighen,
Date
May 19, 1998 (modified) Sep 9, 1996 (created)
|
|