Search

In this article, a workflow for the digitization of complex and extensive periodicals is presented based on the project "Digitization of the Darmstädter Tagblatt" (funded by the German Research Foundation). The project is a collaboration between the University and State Library Darmstadt and the Institute for Linguistics and Literary Studies of the Technical University of Darmstadt. The steps include the creation of images from the original, the digital post-processing up to the machine-readable full text with basic linguistic annotations, the publicly accessible presentation on the Internet and the detailed research for rights holders of individual articles and photographs as stipulated by §§ 61 ff. UrhG ("diligent search"). [ABSTRACT FROM AUTHOR]

This article is based on a Master thesis with the title 'Automatische Indexierung in der sozialwissenschaftlichen Fachinformation. Eine Evaluationsstudie zur maschinellen Erschließung für die Datenbank SOLIS' (Kempf 2012) written within the framework of the postgraduate study program Library and Information Science at Humboldt-Universität zu Berlin at the chair of Information Retrieval. On the basis of the so-called 'Shell Model' (Krause 1996, 2006) for domain-specific content cataloguing it presents evaluation results of an automatic indexing tool for cataloguing of social science research literature. Taking the concrete application scenario formulated by Krause, which suggests that SOLIS-data (Social Science Literature Information System) of less relevance should be indexed automatically, the software MindServer by Recommind was tested in two test series on exactly this data. While in the first test series the system's general settings were tested in the second test series the indexing performance for key and for border areas of the database were compared. For this purpose, sub-discipline-specific versions of the software were built up, which were trained on the basis of corresponding data corpora. The results, evaluated on the basis of intellectually generated comparative data, indicate differences in the quality of indexing for key and for border areas of the database which on the one hand speak against the use of automatic indexing for this area of the database. On the other hand the tests suggest that by building up sub-discipline-specific corpora of training the indexing results could be improved. [ABSTRACT FROM AUTHOR]

DBpedia is a project that among other things extracts RDF statements from articles of Wikipedia. However, it does not exploit full article texts but mainly the so-called infoboxes that contain information that is already structured. As a master thesis at Berlin School of Library and Information Science, wiki2rdf was developed, a tool for rule-based extraction of RDF triples from full unstructured Wikipedia article texts. Extraction is carried out after syntactical parsing using a dependency parser. wiki2rdf has been used for triple extraction from 68820 articles about scientists and humanists (category 'Wissenschaftler') in German Wikipedia. 244563 triples werde extracted. [ABSTRACT FROM AUTHOR]