Authors:
Taoxin Peng
1
and
Calum Mackay
2
Affiliations:
1
Edinburgh Napier University, United Kingdom
;
2
KANA, United Kingdom
Keyword(s):
String Matching, Data Quality, Record Matching, Record Linkage, Data Warehousing.
Related
Ontology
Subjects/Areas/Topics:
Coupling and Integrating Heterogeneous Data Sources
;
Data Warehouses and OLAP
;
Databases and Information Systems Integration
;
Enterprise Application Integration
;
Enterprise Information Systems
;
Performance Evaluation and Benchmarking
Abstract:
Data quality is a key to success for all kinds of businesses that have information applications involved, such as data integration for data warehouses, text and web mining, information retrieval, search engine for web applications, etc. In such applications, matching strings is one of the popular tasks. There are a number of approximate string matching techniques available. However, there is still a problem that remains unanswered: for a given dataset, how to select an appropriate technique and a threshold value required by this technique for the purpose of string matching. To challenge this problem, this paper analyses and evaluates a set of popular token-based string matching techniques on several carefully designed different datasets. A thorough experimental comparison confirms the statement that there is no clear overall best technique. However, some techniques do perform significantly better in some cases. Some suggestions have been presented, which can be used as guidance for r
esearchers and practitioners to select an appropriate string matching technique and a corresponding threshold value for a given dataset.
(More)