Exploratory analyses of structured images: a test on different coding procedures and analysis methods

In order to test the ability of textual coding to depict the features of structured images, a corpus of images of Near-Eastern seals of the late IVth millennium B.C. was studied through different exploratory analysis techniques. Two different coding systems were considered: the classical presence/absence coding of iconographical elements present in the images and a new textual coding, based on a formalised text describing the image. These were submitted to Multiple and Textual Correspondence Analyses. The textual analyses were performed according to two different coding systems, and several choices of the items involved. The results of the different analyses are discussed and compared here. In particular, textual analysis proved effective in substituting the classical coding in the description of the iconographic elements appearing in the images. In addition, it allowed us to broaden the investigation to include aspects of the images (occurrence of fixed sub-patterns and composition) which are beyond the capacities of classical coding. The ability of textual coding to select particular elements, and/or element sequences, to be taken into account in the analyses, was also considered an interesting feature for fine-tuning the analyses to the particular characters of specific corpora. Thus, the use of a formalised text as an intermediate between images and analysis tools proved to be a method worth using, in spite of the special care needed, and some still unsolved difficulties