The program described here implements the first step in voice synthesizing systems, namely, the conversion of common (orthographic) text to its phonemic transcription. There are about 250 linguistically based rules for Polish; all of them are used, but only one example is described in the paper. The rules are based on the assumption that only the continuous left and right context of an orthographic letter determines its phonemic transcription.

This paper is worthwhile reading for application-oriented computational linguists. It contains the whole program (written in Pascal) together with examples. The efficiency of the program is impressive. It seems to be able to run in real-time applications, although the data (i.e., the rules) are separated from the program itself. This again demonstrates that modern structured programming methods are more than viable, even in linguistics. Unfortunately, the implementation of the rule preprocessor is not described in adequate detail, though this can be a problem of almost the same size as the transcription itself.