Corpus Design Principles and Challenges in the COST Action 'Distant Reading for European Literary History, Lecture

In its first part, the lecture outlines the challenges of corpus building for the Romanian ELTeC collection. As shown below, some of them originate in the undisputable linguistic and cultural specificity of Romanian texts, others draw from the post-communist policies concerning the digitization and open-access treatment of cultural heritage: 

a. scarcity of digitized resources from the period 1850-1920, thus a difficult extraction/ checking of metadata; 

b. analysis and automation tools that are still unadjusted to the diachronical particularities of Romanian; 

c. eligibility and composition principles, most of them deducted from Western literary traditions — i.e. book length over 10,000 words, number of editions, 30% canonical works, 10-30% female authors, balanced sampling according to twenty-year time-slots, etc — that are rather inapplicable to Eastern European literary phenomena and institutions; 

d. in the case of Romanian, the slow process of language standardization raises difficult issues concerning clean-up and normalization. For instance, novels published between 1850 and 1865 are printed in a Cyrillic-Latin alphabet that cannot be read by regular OCRs, while novels published after 1865, albeit in Latin alphabet, still have some special glyphs that result in bad OCR output. 

In the second part of the lecture, I go through a few practical solutions that I have tested on the Romanian ELTeC collection: customized digitization (focused on novel subgenres); customized dataset and DOI assignment on Zenodo; support repositories for different text formats on github and zenodo; HTR models for specific prints (such as the ones using the Romanian Transition Alphabet). While they have proven effective for texts belonging to an “emergent”, “peripheral” or “minor” national literature and for a lesser-resourced language such as Romanian, their efficiency should still be checked within a trans-national frame that comprises cultures and languages sharing the same marginal status and that have gone through similar conditions of development.