I am using concordance tools in DTP, and I am curious as to how it determines “similar” words. Does anyone know how those are determined? [I need to justify the use of this in a paper…] Many thanks for any insight.
DEVONthink uses proprietary algorithms to determine the contextual similarity of content of documents, e.g., the Classify and See Also artificial intelligence assistants. Contextual similarity is more complex than similarity of words per se, as it involves frequencies and contextual patterns of words among a document collection.
For example, if I’ve got documents about dogs and select one that doesn’t mention that dogs are canines, See Also may suggest other documents that mention wolves but not dogs, if there are documents about the canine family in my database. In such a case the algorithms will have “bridged” the relationship of dogs to other members of the canine family.
Steven Johnson, in his book Where Good Ideas Come From: The Natural History of Innovation discusses at some length how he uses DEVONthink and See Also in his research and writing.
Thank you very much. I appreciate your time.