A data discovery index for the social sciences

Abstract

This paper describes a novel search index for social and economic research data, one that enables users to search up-to-date references for data holdings in these disciplines. The index can be used for comparative analysis of publication of datasets in different areas of social science. The core of the index is the da|ra registration agency’s database for social and economic data, which contains high-quality searchable metadata from registered data publishers. Research data’s metadata records are harvested from data providers around the world and included in the index. In this paper, we describe the currently available indices on social science datasets and their shortcomings. Next, we describe the motivation behind and the purpose for the data discovery index as a dedicated and curated platform for finding social science research data and gesisDataSearch, its user interface. Further, we explain the harvesting, filtering and indexing procedure and give usage instructions for the dataset index. Lastly, we show that the index is currently the most comprehensive and most accessible collection of social science data descriptions available.

A Digital Library for Research Data and Related Information in the Social Sciences

“…For research data, first, we give access to 6,267 studies from 1945 to 2018 that are collected in the GESIS Data Archive and made accessible through the GESIS data catalog. Second, we include about 77,432 studies from international players such as ICPSR, DANS (Data Archiving and Networked Services Netherlands) or the UK Data Archive which have been harvested and made accessible in the da|raSearchNet project [18]. This gives the so-far best overview of research data in the field of social sciences.…”

“…( In the search interface, we combined the support needs (4-7) into the category "Instruments and Tools". Literature (1), research data (2), and variables and questions (3) got their own categories because of their expressed importance and the 18 www.gesis.org/IIRpanel comprehensive data in each category. So far networking and illustrative material are not included in the integrated search system.…”

“…A few of them worked at our institute but were not involved in the development of the integrated search system. Most of the participants were external social science researchers recruited by mailing lists, the IIRpanel18 , or poster announcement at the University of Cologne, Düsseldorf, and Mannheim.…”

Genuine Information Needs of Social Scientists Looking for Data

“…Existing vocabulary standards need to be extended with the help of the community, tools must exist for documentation, and published metadata records need to be available in search indices. For example, Krämer et al (2018) assessed the availability and quality of 58 social science data providers' metadata. The adoption of the DDI metadata standard was low compared to the more generic Dublin Core standard, and metadata fields were often left unused.…”

Learning Domain‐specific Semantic Representation from Weakly Supervised Data to Improve Research Dataset Retrieval

“…To evaluate the proposed research dataset retrieval model, we generated a set with 104,683 research datasets collected from 13 well‐known data repositories related to humanities and social sciences, including Harvard Dataverse, Inter‐university Consortium for Political and Social Research (ICPSR), UK Data Service, Research Data Australia, the Council of European Social Science Data Archives, the Digital Archaeological Record, et al The Selected repositories come from frequently recommended repositories of data journals, top universities, and their libraries. As the metadata format of each repository is quite different, for example, the metadata format of Research Data Australia is the Registry Interchange Format ‐ Collections and Services (RIF‐CS), and the metadata format of ICPSR is Data Documentation Initiative (DDI), to search datasets from different repositories in a unified manner, referencing to existing dataset retrieval systems such as gesisDataSearch (Krämer et al, 2018) and Federated Research Data Repository (Garnett et al, 2017), we mapped each repository’s metadata to Dublin Core (DC) according to manually made rules. In total, we created a dataset retrieval system that indexed 104,683 datasets with 15 DC metadata fields.…”