View GERMANA (public)
GERMANA is the result of digitising and annotating a 764-page Spanish manuscript
- Handwriting-Recognition Historical-Documents
- Attribute Types
.tar.gz (1.1 GB)
Files are converted on demand and the process can take up to a minute. Please wait until download begins.
You can edit this item to add more meta information and make use of the site's premium features.
- Original Data Format
- Version mldata
- Data (first 10 data points)
(Zipped) TAR archive GERMANA.0.tgz, GERMANA.1.tgz, GERMANA.2.tgz, GERMANA.3.tgz, GERMANA.4.tgz, GERMANA.5.tgz, GERMANA.6.tgz, GERMANA.7.tgz, COPYING
GERMANA is the result of digitising and annotating a 764-page Spanish manuscript entitled “Noticias y documentos relativos a Doña Germana de Foix, ́última Reina de Aragón" and written in 1891 by Vicent Salvador, the Cruïles marquis.
It has approximately 21K text lines manually marked and transcribed by palaeography experts. It is a single-author book on a limited-domain topic: the life of Germana de Foix (1488-1538), niece of King Louis XII of France and second wife of Ferdinand the Catholic of Aragon. The original manuscript was well-preserved and most pages only contain nearly calligraphed text written on ruled sheets of well-separated lines. Moreover, the manuscript comprises about 217K running words from a vocabulary of 30K words which, apparently, is a reasonable amount of data for single-author handwriting and language modelling. It goes without saying that text line extraction and off-line handwriting recognition on GERMANA is not, by contrast, particularly easy.
GERMANA has typical characteristics of historical documents that make things difficult: spots, writing from the verso appearing on the recto, unusual characters and words, etc. Also, the manuscript includes many notes and appended documents that are written in languages different from Spanish, namely Catalan, French and Latin. GERMANA entails an appropriate trade-off between task complexity and amount of data. It is the first publicly available database for handwriting research, mostly written in Spanish and comparable in size to standard databases such as IAM. Due to its sequential book structure, it is also well-suited for realistic assessment of interactive handwriting recognition systems. Moreover, it can be used as well to test approaches for language identification and adaption from singleauthor handwriting.
A new handwritten text database, GERMANA, is presented to facilitate empirical comparison of different approaches to text line extraction and off-line handwriting recognition. GERMANA is the result of digitising and annotating a 764-page Spanish manuscript from 1891, in which most pages only contain nearly calligraphed text written on ruled sheets of well-separated lines. To our knowledge, it is the first publicly available database for handwriting research, mostly written in Spanish and comparable in size to standard databases. Due to its sequential book structure, it is also well-suited for realistic assessment of interactive handwriting recognition systems. To provide baseline results for reference in future studies, empirical results are also reported, using standard techniques and tools for preprocessing, feature extraction, HMM-based image modelling, and language modelling.
- Data Source
- The manuscript is an annotated version of the scanned version at the "Biblioteca Valenciana Digital".
- Measurement Details
The typical task in RODRIGO is transcribe it from the begining to the end. Transcription quality is expressed in terms of Word Error Error (WER) rate.
- Usage Scenario
GERMANA annotation includes line transcriptions, as well as their corresponding baselines, and text block coordinates. This annotation can be used to test methods for: document layout analysis, line detection, and handwritten text recognition.
No one has posted any comments yet. Perhaps you would like to be the first?
Leave a commentTo post a comment, please sign in.
This item was downloaded 449 times and viewed 313 times.
No Tasks yet on dataset GERMANASubmit a new Task for this Data item
We are acting in good faith to make datasets submitted for the use of the scientific community available to everybody, but if you are a copyright holder and would like us to remove a dataset please inform us and we will do it as soon as possible.
This project is supported by PASCAL (Pattern Analysis, Statistical Modelling and Computational Learning)