Exploration of contextual constraints for character pre-classification

We present strategies and results for identifying the symbol type (lower-case, upper-case, digit, and punctu-ation or special symbols) of every character in a text document by using various kinds of information from neighboring characters. In the expectation of reason-able word and character segmentation for shape cluster-ing, we designed several type recognition methods that depend on cluster n-grams, shape codes, and within-word context. On an ASCII test corpus of 925 arti-cles that simulates perfect image-level processing, these methods achieve a substantial improvement over default assignment of all characters to lower case. 1