An algorithm for analyzing cloze test results

Abstract

We present an algorithm for the automatic processing of cloze test results that are traditionally analyzed and adjusted by hand. The algorithm accurately includes inflected and derived forms of the target word as correct responses, as well as minor spelling mistakes and typographical errors. We provide a test of the algorithm against the coding of one expert human observer and find that the algorithm performs with equal accuracy. We further examine the utility of the algorithm for the analysis of data involving word identification with typed responses. The routines are implemented in R-software and are freely available.