University of Warsaw
Biblioteka Cyfrowa KLF UW (Digital Library of the Formal Linguistics Department at the University of Warsaw)Not a member yet
403 research outputs found
Sort by
Skanowane teksty jako korpusy
No full textScanned texts as corpora
A modification of the Poliqarp corpus search tool is described, which is oriented towards searching scanned texts with dirty OCR (i.e. the fully automatic Optical Character Recognition without any proof-reading). This search tool operates since December 2009 and is available at http://wbl.klf.uw.edu.pl/. The two-level regular expressions, which can be used in the queries, allow, at least in principle, to circumvent the OCR errors.
The crucial property of the search engine is its ability to highlight the hits on the original scans stored in the DjVu format. Although the feature is not original, as it has been used for the first time for the Century Dictionary and later for Jamieson’s Etymological Dictionary of the Scottish Language, it is substantially augmented by allowing so called graphical concordances and providing a convenient way to bookmark the hits.
Our system handles now four dictionaries, with the total size of over 40 000 pages. It is expected that in the near future other texts will be added to the system
Narzędzia dygitalizacji tekstów na potrzeby badań filologicznych
No full textSprawozdanie merytoryczne z grantu MNSzWiT nr N N519 384036, 13.05.2009-12.05.2012
Scanned publications in digital libraries: new Open Source DjVu tools
No full textThe DjVu technology is described by its authors as "an image
compression technique, a document format, and a software platform for
delivering documents images over the Internet"; according to the
recent statistics, about 80% of documents stored in Polish digital
libraries is in this format. Besides the commercial software
supporting this technology there is also the DjVuLibre suit of Open
Source tools and utilities, developed by the technology creators.
In the presentations another Open Source suit of programs will be
discussed. It consist of two sets.
The first set contains some programs for creation and improvement of
DjVu documents including the results of Optical Character
Recogniton. A typical OCR program outputs the results as a PDF
"sandwich" document containg text under image (although since version
11 ABBY FineReader can save directly the output as a DjVu files, the
output in the PDF form contains more information). The pdf2djvu
program conceived by Jakub Wilk (http://jwilk.net/software/pdf2djvu)
convert the PDF files into DjVu preserving all the features
(e.g. outlines) which are representable in the latter format.
The purpose of another program, also conceived by Jakub Wilk
(http://jwilk.net/software/didjvu), is the conversion of graphic files
into the DjVu documents consisting of foreground (the printed text),
mask and background layers (e.g. illustrations). Such separation not
only allows to achieve a high compression ratio, but also improves the
quality of OCR results which should operate only on foreground or
mask.
The third program named, for the historical reasons, ocrodjvu
(http://jwilk.net/software/ocrodjvu) is a wrapper for several Open
Source OCR programs including Tesseract, which achieves quality
comparable with commercial systems
(cf. e.g. a test results).
The second set of programs concerns the delivery of DjVu documents to
the users. It consist of a search engine server and two kind of
clients: marasca installable as a WWW site and djview4poliqarp, a
standalone client installable on a user computer. As the server is
based on the Poliqarp corpus tool, the whole set is called just
Poliqarp for DjVu. The author of djview4poliqarp is Michał Rudolf, the
rest of the system was created by Jakub Wilk.
The tools has been developed in the framework of the project directed
by the present author, the results are available on the principle of
GNU General Public License
