On partitioning a dictionary for visual text recognition

摘要

In this paper we examine some of the possible ways of partitioning a dictionary in the context of visual text recognition. We have considered approximately 10,000 of the most frequently used words for our analysis and comparison. We have used word-length, word-envelope, character combination and a combination of these as the basis of partitioning the dictionary. These features do not require additional computation as they are mostly available as a by-product of the on-going process of character recognition. In all, 11 different schemes have been examined and evaluated with respect to real-life problems faced in text recognition.