Data Normalization Procedures on Decomposed MARC 21 Records
Edward Kim, William E. Moen · 2001
In this document, the authors present some aspects of data normalization of the decomposed records to improve the results of analysis. The data normalization processes use pattern-matching techniques to eliminate and/or generalize anomalous characters and terms. Since the unit of analysis in preparing the test dataset of 400,000 MARC 21 records is a word, there was a need for data normalization to provide reliability in the subsequent analysis.