Nederlab: Towards a Single Portal and Research Environment for Diachronic Dutch Text Corpora

Nederlab: Towards a Single Portal and Research Environment for Diachronic Dutch Text Corpora Hennie Brugman1, Martin Reynaert34, Nicoline van der Sijs12, Ren´e van Stipriaan1 , Erik Tjong Kim Sang1, Antal van den Bosch2 Meertens Institute Amsterdam1, CLS2 – CLST3 / Radboud University Nijmegen, TiCC / Tilburg University4, hennie.brugman@meertens.knaw.nl, reynaert@uvt.nl, post@nicolinevdsijs.nl, stipriaan@planet.nl, erik.tjong.kim.sang@meertens.knaw.nl, a.vandenbosch@let.ru.nl Abstract The Nederlab project aims to bring together all digitized texts relevant to the Dutch national heritage, the history of the Dutch language and culture (circa 800 – present) in one user friendly and tool enriched open access web interface. This paper describes Nederlab halfway through the project period and discusses the collections incorporated, back-office processes, system back-end as well as the Nederlab Research Portal end-user web application. Keywords: Dutch, Diachronic Corpora, Online Portal, Nederlab 1. Introduction Many digital Dutch text collections, from the eighth cen- tury to the present day, are available for digital humanities researchers. These texts reflect the dynamic development of Dutch language and culture. However, these collections typically are hosted by different institutions, are described with different metadata, and cannot be searched simultane- ously. Advanced software tools necessary for the analysis of the texts in these collections are available, but are in most cases specific to particular goals and formats, rather than being generally applicable. The Nederlab project is building a comprehensive research corpus that brings together collections from various sources in the Netherlands and Flanders. It also brings together rel- evant tools for processing, searching and analyzing these data, in one virtual research environment1 that primarily supports historians, literary scholars, and linguists. The specific goal of Nederlab is to discover patterns of change over time and space. Collections that are aggregated and integrated into Neder- lab are processed using a carefully designed Nederlab col- lection pipeline (see Section 4.). This pipeline is applied to a number of basic resource types (several types of ti- tles, persons/authors). It deals with metadata, text content, some vocabularies (e.g. genres, places) and linguistic an- notations. It first aggregates data, then maps and converts it to a limited number of pivot formats, it identifies and links related resources. The pipeline involves an editorial team that is responsible for data selection, quality assessment and manual and semi-automatic data curation and interlink- ing. For texts that originate from Optical Character Recog- nition automatic spelling correction, OCR post-correction and text normalization is applied. To all texts we apply sentence splitting, tokenization, lemmatization, part of speech tagging, and named entity recognition, resulting in massive amounts of linguistic an- notations. This uniform annotation treatment enables us to obtain the necessary statistics for word tokens and types, lemmata, PoS tags and entities across all incorporated cor- 1Online at http://www.nederlab.nl/ pora. Federated Search (Stehouwer et al., 2012) across non- uniformly annotated corpora simply cannot provide this. Metadata, text and linguistic annotations are indexed in one large, powerful search index. Finally, a broker orchestrates searching and other data services, and provides a uniform web-based access point to all Nederlab collections for end- user tools. The …

Content truncated to 500 words