Crowdsourcing an OCR Gold Standard for a German and French Heritage Corpus

Crowdsourcing an OCR Gold Standard for a German and French Heritage Corpus Simon Clematide, Lenz Furrer, Martin Volk Institute of Computational Linguistics University of Zurich, Switzerland {simon.clematide,furrer,volk}@cl.uzh.ch Abstract Crowdsourcing approaches for post-correction of OCR output (Optical Character Recognition) have been successfully applied to several historic text collections. We report on our crowd-correction platform Kokos, which we built to improve the OCR quality of the digitized yearbooks of the Swiss Alpine Club (SAC) from the 19th century. This multilingual heritage corpus consists of Alpine texts mainly written in German and French, all typeset in Antiqua font. Finding and engaging volunteers for correcting large amounts of pages into high quality text requires a carefully designed user interface, an easy-to-use workflow, and continuous efforts for keeping the participants motivated. More than 180,000 characters on about 21,000 pages were corrected by volunteers in about 7 months, achieving an OCR gold standard with a systematically evaluated accuracy of 99.7 % on the word level. The crowdsourced OCR gold standard and the corresponding original OCR recognition results from Abby FineReader 7 for each page are available as a resource. Additionally, the scanned images (300 dpi) of all pages are included in order to facilitate tests with other OCR software. Keywords: Optical Character Recognition, Crowdsourcing, Cultural Heritage Corpora 1. Introduction Crowdsourcing approaches for post-correction of OCR output (Optical Character Recognition) have been suc- cessfully applied to several historic text collections. We report on our crowd-correction platform Kokos, which we built to improve the OCR quality of the dig- itized yearbooks of the Swiss Alpine Club (SAC) from the 19th century. This multilingual heritage corpus consists of Alpine texts mainly written in German and French, all typeset in Antiqua font. Finding and engaging volunteers for correcting large amounts of pages into high quality text requires a care- fully designed user interface, an easy-to-use workflow, and continuous efforts for keeping the participants mo- tivated. The scanned images, the uncorrected output of a stan- dard OCR software and the high-quality text corrected by our crowd build a valuable resource.1 It can be used for extracting heritage lexicons covering 19th century German in particular, or for training as well as testing automatic OCR error correction systems. In the following section, we introduce our multilin- gual corpus and describe the process of its digitization. We report on our efforts in building and maintaining a crowd-correction platform and compare them to other work in the field. In Section 3, we analyze and evalu- ate the corrections performed by the volunteer collab- orators. The released resource is described in the last section. 2. Materials and Methods 2.1. Corpus Data In the Text+Berg project2 we digitized the yearbooks of the SAC from 1864 until today (henceforth SAC 1http://pub.cl.uzh.ch/purl/OCR19thSAC 2http://textberg.ch corpus) for building a multilingual heritage corpus of Alpine texts (Göhring and Volk, 2011). In this paper we focus on the yearbooks of the 19th century. Those yearbooks from 1864 to 1899 amount to 21,247 pages (including tables of content and index pages) with around 304,000 …

Content truncated to 500 words