This paper describes a methodology to perform automatically a standard assessment of acoustic phonetic decoders. Results such as confusion matrix, deletion or insertion matrices are computed to provide a global overview of the system assessment. This method relies on the dynamic programming algorithm. It matches the outputs of the decoder with a standard transcription of the spoken sentence and its most common utterances derived with phonological rules. The errors made by the assesser are computed as errors of the decoder. In order to minimize this drawback, we guide the matching process by running an adaptation of the assessment algorithm to the decoder. The assessment of APHODEX, the acoustic phonetic decoder developed in CRIN, is performed and results are discussed.