The GERMANA Database

Language Identification for Interactive Handwriting Transcription of Multilingual Documents

“…As indicated in the introduction, the experiments reported here were carried out on a multilingual, single-author manuscript from 1891 known as GERMANA database [3]. Our main goal is to empirically compare the three language identification techniques described in the preceding section.…”

“…A good example of multilingual document is the GERMANA database [3]. GERMANA is the result of digitizing and annotating a 764-page, single-author Spanish manuscript from 1891, solely written in Spanish up to page 180, but then also written in five other languages, especially Catalan and Latin.…”

Character-Based Handwritten Text Recognition of Multilingual Documents

“…On the other hand, punctuation signs ("," after "Vidal" and "Reina"), are successfully recognized in the character-based approach, whereas, the word-based approach failed to recognize this signs due to its scarcity in the training dataset. In past works [6], we only dealt with GERMANA first part, where we reported a performance of 34.51% of WER, in this same partition, the character-based system obtained a performance of 12.12% WER. …”

“…Experiments were performed in the GERMANA database [6]. GERMANA is a single-author manuscript from 1891, which contains 764 pages written in up to six different languages.…”

“…A good example of multilingual document is the GERMANA database [6]. GERMANA is the result of digitizing and annotating a 764-page, single-author manuscript from 1891, written in Spanish up to page 180, but then also written in five other languages, mainly in Catalan and Latin.…”