Offline Word Spotting in Handwritten Documents

The digitization of written human knowledge into string data has reached up to but not beyond the recognition of typeset text. This means that vast libraries of handwritten, cursive documents must be indexed and transcribed by a human—a prohibitively laborious task. This paper explores an existing technique developed in [1] and [12] for the offline indexation of historical handwritten documents. The algorithm clusters segmented word images using Dynamic Time Warping (DTW) which compares sets of time-dependent feature vectors from two images. By clustering words in an unknown document, a human only has to label the word clusters, significantly reducing the human’s workload. Currently the algorithm achieves a 40.8% classification rate using only two features. Other research from the community suggests that a classification rate of between 65% and 72% is achievable with additional features and further refinement to the algorithm.