Search

In this paper, we describe our effort to create a new corpus for the evaluation of detecting and linking so-called survey variables in social science publications (e.g., ”Do you believe in Heaven?”). The task is to recognize survey variable mentions in a given text, disambiguate them, and link them to the corresponding variable within a knowledge base. Since there are generally hundreds of candidates to link to and due to the wide variety of forms they can take, this is a challenging task within NLP. The contribution of our work is the first gold standard corpus for the variable detection and linking task. We describe the annotation guidelines and the annotation process. The produced corpus is multilingual – German and English – and includes manually curated word and phrase alignments. Moreover, it includes text samples that could not be assigned to any variables, denoted as negative examples. Based on the new dataset, we conduct an evaluation of several state-of-the-art text classification and textual similarity methods. The annotated corpus is made available along with an open-source baseline system for variable mention identification and linking.

The ability to browse an information space in a structured way by exploiting similarities and dissimilarities between information objects is crucial for knowledge discovery. Knowledge maps use visualizations to gain insights into the structure of large-scale information spaces, but are still far away from being applicable for searching. The paper proposes a use case for enhancing search term recommendations by heat map visualizations of co-word relationships taken from indexing vocabulary. By contrasting areas of different “heat” the user is enabled to indicate mainstream areas of the field in question more easily. (author's abstract)

The dramatic increase in information on the World Wide Web makes it more difficult for web users using web search engines to effectively satisfy their information needs The users' lack of knowledge regarding the searched topics creates a complicated problem when formulating an effective query. Query expansion can play an essential role in overcoming such a deficit. However, because they lack sufficient knowledge about the searched topics, users sometimes find it difficult to evaluate the relatedness of the system's automatically expanded terms. This problem arises mostly in domain-specific areas, e.g Social Science. In this paper, we expand the queries based on a structured, open knowledge resource on the Web (Wikipedia). We link the entities from a domain-specific corpus (qualitative journals in social science) to Wikipedia entities. With Wikipedia serving as background knowledge, we help users with their selections by providing the most likely related terms. Furthermore, users are provided with contextual information that describes each expanded term in order to give users a clearer idea about the meaning of each expanded term. By utilizing 10 test queries chosen by experts who also evaluated the results, we compared the results of using Wikipedia with the results achieved by using a qualitative journal (author's abstract)

This paper is about a better understanding on the structure and dynamics of science and the usage of these insights for compensating the typical problems that arises in metadata-driven Digital Libraries. Three science model driven retrieval services are presented: co-word analysis based query expansion, re-ranking via Bradfordizing and author centrality. The services are evaluated with relevance assessments from which two important implications emerge: (1) precision values of the retrieval service are the same or better than the tf-idf retrieval baseline and (2) each service retrieved a disjoint set of documents. The different services each favor quite other - but still relevant - documents than pure term-frequency based rankings. The proposed models and derived retrieval services therefore open up new viewpoints on the scientific knowledge space and provide an alternative framework to structure scholarly information systems., 8 pages, 4 figures, Cologne Conference on Interoperability and Semantics in Knowledge Organization

"Der Arbeitsbericht informiert über Entwicklungen am IZ, die darauf abzielen, Wissen über das Interaktionsgeschehen in wissenschaftlichen Communities und den sozialen Status ihrer Akteure für das Retrieval auszunutzen. Grundlage hierfür sind soziale Netzwerke, die sich durch Kooperation der wissenschaftlichen Akteure konstituieren und in den Dokumenten der Datenbasis z.B. als Koautorbeziehungen repräsentiert sind (Autorennetzwerke). Die in dem Bericht beschriebenen Studien zur Small-World-Topologie von Autorennetzwerken zeigen, dass diese Netzwerke ein erhebliches Potential für Informationssysteme haben. Kernansatz der beschriebenen Retrievalmodelle ist die Suche nach Experten und das Ranking von Dokumenten auf der Basis der Zentralität von Autoren in Autorennetzwerken." (Autorenreferat)

"Die Sonderfördermaßnahme CARMEN ('Content Analysis, Retrieval and Metadata: Effective Networking') zielte im Rahmen des vom BMB+F geförderten Programms GLOBAL INFO unter anderem auf eine (semantische) Verbesserung der Ausweitung von Recherchen in Fachdatenbanken zu Internet-Recherchen und umgekehrt ab. Die dafür gewählten Konzepte zur Behandlung semantischer Heterogenität wurden entwickelt, implementiert und evaluiert. Der vorliegende Bericht beschreibt die Arbeiten und Ergebnisse der Projektpartner dieses Arbeitspakets (Fachbereich Mathematik/Informatik der Universität Osnabrück sowie das Informationszentrum Sozialwissenschaften Bonn)." (Autorenreferat)

"Nowadays, users of information services are faced with highly decentralised, heterogeneous document sources with different content analysis. Semantic heterogeneity occurs e.g. when resources using different systems for content description are searched using a single query system. This report describes several approaches of handling semantic heterogeneity used in projects of the German Social Science Information Centre." (author's abstract)

"Kommerzielle Textretrievalsysteme basieren heute im wesentlichen auf intellektuell und/oder automatisch ermittelten Deskriptoren, die mit oder ohne zusätzliche Thesaurusrelationen mit Hilfe der Booleschen Algebra recherchiert werden. Dem stehen die quantitativ-statistischen Verfahren gegenüber, die zu nach Relevanz geordneten Ergebnislisten führen (ranking). Sie setzten sich in den letzten Jahren verstärkt auch im kommerziellen Bereich durch. Im vorliegenden Arbeitsbericht wird der Frage nach dem adäquatesten Inhaltserschließungskonzept für ein Informationssystem nachgegangen. Es werden die Vor- und Nachteile der intellektuellen Inhaltserschließung und der automatischen Indexierung im Kontext von Booleschen und quantitativ-statistischen Recherchesystemen ganz generell diskutiert. Desweiteren erfolgt eine Evaluierung der informationstechnologischen Grundlagen für ein modernes Informationssystem." (Autorenreferat)