April 30, 2005

Ambiguity

In reference to one of my recent posts, where the Bibliothèque Nationale de France was abbreviated as "BNF", Mike Albaugh complained that

BNF still means Backus/Naur Form to me, at first glance. (or Normal :-)

OK, me too. But in interpreting abbreviations, you've got to consider the context, as an Australian traffic court judge recently explained to the author of a clever new defense against a speeding ticket:

An early contender for the 2005 Nice Try But No Cigar award goes to Carl Ross La Riviere. Yesterday in the District Court La Riviere was avidly defending himself against a speeding charge, even though the Crown appeared to have irrefutable proof: a photo of his car travelling 56 "KPH" in a 40 "KPH" zone.

La Riviere presented the court with the National Measurement Act and Regulations to prove these initials did not mean what everyone thinks they do. His argument: K stands for Kelvin, a measurement of thermodynamic temperature; P for Poise, a measurement of viscosity; and H for Henry, which measures electric inductance.

Entertained but unmoved, Judge Megan Latham agreed the abbreviation might be illogical or incorrect, but had to be seen in context, which clearly suggested it related to a vehicle speed and meant kilometres per hour. La Riviere's sentence - a good behaviour bond - was upheld.

Mike also wondered about the relationship between plan calcul and Plankalkül,

...the names given to the French national computing initiative (circa 1966) and Konrad Zuse's proposed algorithmic language (or notation :-) (circa 1946).

When I pointed him to the entry in the French Jargon File that says there's "aucun rapport" ("no relationship"), Miked answered:

Well, that's just what they would say, isn't it? :-)

and (returning to BNF) suggested that it's

Time to revive SAFEBAGEL (Scientists Against Far-out, Extensive, Burdensome Acronyms Getting Entrenched in Language).

Scientists are the worst offenders. A couple of years ago, I wrote a little program to find acronyms in the MEDLINE corpus. There are lots of them -- my not-very-smart program found more than 78,000 distinct acronym/definition pairings, many of which occurred many times. Thus GM-CSF was defined 2,401 times as "granulocyte-macrophage colony-stimulating factor", but was also defined by 150 other strings. In this case, these are basically all variant forms of the same term (including a shocking number of typos -- it seems that biomedical journals are not always very well copyedited) -- see this page for the complete list of variants, each preceded by the number of times my program found it as a definition for GM-CSF in MEDLINE.

There were also plenty of acronyms whose definitions were not just different versions of the same term. For example, ABA was variously abscisic acid, Agaricus bisporus agglutinin, aminoalkyl-iodobenzamides, aminobenzamide, aminobenzanthrone, anti-biotin antibody and azobenzenearsonate. With a bit of extra context, ABA could be part of I-ABA (Iodo-4-aminobenzyl adenosine), PABA (para-aminobenzoic acid or pyridylamino butylamine), and hundreds of other things.

For most Americans, ABA is the American Bar Association. Down under, it could be the Australian Broadcasting Authority. For others, it might be the Association for Behavior Analysis or the American Board of Anesthesiology or the Antiquarian Booksellers Association. It shows my age that for me, ABA will always be first and foremost the American Basketball Association.

Life, like language, is ambiguous. Without the effect of context, referential communication would hardly ever succeed.

Posted by Mark Liberman at April 30, 2005 11:51 AM