We explore the relationship between the lexical structure and the syntactical structure of numerical expressions in number transcoding from the oral verbal format to the Arabic digital format. The experimental setup included asking six to eight-year old Spanish-speaking children attending elementary school to write in Arabic format a set of dictated numerals. The method of analysis includes the construction of a relational representation of children's production and the use of clustering techniques to identify patterns. The model relates children and dictated numerals by children's accomplishment and generates a subsidiary similitude relation between dictated numerals with patterns that show differentiated structures. We find that the presence or absence of a verbal expression for the Hundred position digit in the dictated numeral marks one of the structures. The second structure comes from the role of two digit numbers (e.g. 20 or 34): homogeneous in the Decade position and heterogeneous in the Thousand position. We interpret these results as consistent with the semantic-lexical internal number representation model by R.J.D. Power and M.F. Dal Martello, The dictation of Italian numerals, Language and Cognitive Processes, 5, 237-254; 1990.