Evaluation of the Effect of Improper Segmentation on Word Spotting
This addresses a practical issue for researchers in historical document analysis by highlighting the gap between ideal and real-world conditions, though it is incremental as it focuses on evaluation rather than new methods.
The paper tackles the problem of evaluating word spotting methods under imperfect segmentation by proposing an experimental framework to quantify segmentation effects, finding that performance degrades significantly with distortion, e.g., retrieval accuracy drops by up to 30% on benchmark datasets.
Word spotting is an important recognition task in historical document analysis. In most cases methods are developed and evaluated assuming perfect word segmentations. In this paper we propose an experimental framework to quantify the effect of goodness of word segmentation has on the performance achieved by word spotting methods in identical unbiased conditions. The framework consists of generating systematic distortions on segmentation and retrieving the original queries from the distorted dataset. We apply the framework on the George Washington and Barcelona Marriage Dataset and on several established and state-of-the-art methods. The experiments allow for an estimate of the end-to-end performance of word spotting methods.