Relevance

From Memography

In the field of library science, precision and recall have clear mathematical definitions.

precision = the number of relevant documents returned/the number of documents returned.

recall = the number of relevant documents returned/the total number of relevant documents (in the universe searched).

But what is relevant? The concept of relevance has always been highly subjective. Ever since Cyril Cleverdon and his colleagues at Cranfield developed their metrics for precision and recall in the late 'fifties there has been continuing debate about "relevance." (Svenonius, p. 22.) [1]

The modern debate is about using "precoordination" (using subject heading languages) as opposed to just searching keywords, even with Boolean coordination (Mortimer Taube, 1951).

Svenonius covers the pros and cons on p.188ff, in terms of specificity, precision, contextuality, and suggestibility.

Memography offers the potential of developing the most specific precoordinated terms ever, and off-loading the heavy lifting onto the meme owners and meme taggers (a/k/a catalogers, indexers, classifiers).

When memography makes the relevance criterion simply that the returned page has been tagged with the meme ID, we put the responsibility on the meme taggers, who control the contents of the aboutness page and, as we will see, control over which pages are retrieved that contain their meme ID.

Because this definition may do violence to the current usage of relevance, we suggest calling it "relevence," with deference to Jacques Derrida, whose "differance" made such a difference.

See the relevence of misspellings.

[1] Svenonius, Elaine, The Intellectual Foundation of Information Organization, MIT Press, 2000.