A Historical Document Handwriting Transcription End-to-end System

摘要

To provide access to the contents of the document collections that are being digitized, transcription is required. Unfortunately manual transcription is generally too expensive and, in most cases, current automatic techniques fail to provide the required level of accuracy. An alternative that can speed up and lower the cost of this process is the use of computer assisted, interactive techniques. These techniques work at line-level thus the transcription task assumes that the page images have been correctly decomposed into the relevant text line images. In this paper we present an end-to-end system that takes as input a page image and provides a fully correct transcript with the help of user interaction. The system automatically performs the text block and text line detection to be fed into the interactive computer assisted transcription. Experiments carried out show that the expected amount of user effort needed to produce perfect transcripts, can be reduced by using the proposed end-to-end system.