Modelling the archaeologist’s thinking for the automatic classification of Uruk/Jamdat Nasr seal images

The classification of Jamdat/Nasr period seals based on the contents of the images engraved on them needs some special coding and data analysis method. In previous studies (Camiz et al. 1998) a Levenshtein–like distance was developed as association measure among the images syntactical structure. The aim of the paper is to improve the previous studies and to describe a new method, based on the archaeologist’s thought. In order to compute a distance among the two symbolic sequences that describe the syntactic structure of the images, a decision tree is taken in account, that imposes different choices and adapts the distance computations according to the followed branches. In this way, both the coding and the weighting system are not affected, but the method seems more apt to take into account the differences in the general image structure.