Structure, Effectiveness, and Uses of the citation Identifier

A computer program for automatic identification of "full-form" case citations in legal literature (e. 2D 117) has been developed at the University of Pittsburgh and is now operational. The level of performance of this program known as "The Citation Identifier" is high. In a recent computer run, ~ae Citation Identifier scanned the full texts of 191 randomly selected decisions of U.S. Court of Appeals (some 400, 000 words of running text) and located correctly 2,220 full-form citations out of a total of 2, 227 (that is better than 99% of the total). Only seven misses and three false drops occurred. Of 2, 220 full-form citations which were located correctly , 1944 (87%) were identified perfectly. In addition, there were 276 partial identifications containing two types of errors : (1) partial identifications in which some citation terms were mistakenly lopped off by the program (so-called "short hits"), and (2) partial identifications which contained words that were improperly included in the citations (so-called "long hits"). Both types of errors are for the most part easily cor-rectable and can be largely eliminated by suitable changes in the program.-1-The Citation Identifier operates rather rapidly. In a recent test run, the total time required to process some 400, 000 running words of t~xt was approximately fifteen and a half minutes. This speed could be further increased by suitable changes in the cc~uter program. An extension of the Citation Identifier to reduced-form citations (e.g., "The Geddes decision," "the Southland Industries ease") is no~ in preparation.