The ESPOSALLES database: An ancient marriage license corpus for off-line handwriting recognition

Abstract

Historical records of daily activities provide intriguing insights into the life of our ancestors, useful for demography studies and genealogical research. Automatic processing of historical documents, however, has mostly been focused on single works of literature and less on social records, which tend to have a distinct layout, structure, and vocabulary. Such information is usually collected by expert demographers that devote a lot of time to manually transcribe them. To support research in automatic handwriting recognition for historical documents containing social records, this paper presents a new database compiled from a marriage license books collection. Marriage license books are documents that were used for centuries by ecclesiastical institutions to register marriage licenses. Books from this collection are handwritten and span nearly half a millennium until the beginning of the 20th century. In addition, a study is presented about the capability of state-of-the-art handwritten text recognition systems, when applied to the presented database. Baseline results are reported for reference in future studies.

On the Modification of Binarization Algorithms to Retain Grayscale Information for Handwritten Text Recognition

“…In the pre-processing module, the pages are first divided into line images as explained in [8]. Given that it is quite common for handwritten documents to suffer from degradation problems it is necessary to first apply appropriate filtering methods to remove the background, noise, improve the quality of the image and to make the documents more legible.…”

“…Three different methods have been tested. The first one is the background estimation and subtraction method used in [8], the second one is the classical Sauvola thresholding [10], and the final one is the proposed method described in Section 3. In Section 5 we can see the different results obtained with the three methods.…”

“…Finally the size is normalized separately for each line. A more detailed description of this process is found in [8]. Then, each pre-processed line image is represented as a sequence of feature vectors representing gray levels and gradients [12].…”

“…In this paper the experiments have been carried out on the publicly available ESPOSALLES 1 database [8]. Most specifically, we used the LICENSES part, which was compiled from a handwritten marriage license book conserved at the Archives of the Cathedral of Barcelona.…”

“…The complete annotation of LICENSES contains around 60,000 running words from a lexicon of around 3,500 different words. More information can be found at [8].…”