Maintaining an Online Bibliographical Database: The Problem of Data Quality

CiteSeer and Google-Scholar are huge digital libraries which provide access to (computer-)science publications. Both collections are operated like specialized search engines, they crawl the web with little human intervention and analyse the documents to classifiy them and to extract some metadata from the full texts. On the other hand there are traditional bibliographic data bases like INSPEC for engineering and PubMed for medicine. For the field of computer science the DBLP service evolved from a small specialized bibliography to a digital library covering most subfields of computer science. The collections of the second group are maintained with massive human effort. On the long term this investment is only justified if data quality of the manually maintained collections remains much higher than that of the search engine style collections. In this paper we discuss management and algorithmic issues of data quality. We focus on the special problem of person names.

• Type: Conference
• Year: 2006
• Where: FEGC
• Authors: Michael Ley, Patrick Reuther
• Keywords: Biometrics, FEGC 2006, Huge Digital Libraries, Little Human Intervention, Search Engine