Record Linkage For Character-Based Surnames: Evidence from Chinese Exclusion.

Area of Science:

• Historical Demography
• Data Science
• Computational Linguistics

Background:

• Traditional record linkage methods struggle with unique challenges in historical Chinese datasets, such as naming conventions and data errors.
• Existing approaches lack effectiveness for non-Romanized scripts and culturally specific data structures.
• Improving data linkage is crucial for accurate historical population research.

Purpose of the Study:

• To develop and evaluate a novel pre-processing technique for enhancing record linkage in historical Chinese population data.
• To address the limitations of current matching algorithms when applied to Chinese naming conventions and enumeration errors.
• To increase both the rate and accuracy of record linkage for historical datasets.

Main Methods:

• A three-step pre-processing procedure was developed specifically for Chinese character-based data.
• The technique involves data cleaning and standardization tailored to historical Chinese naming practices.
• The proposed method was tested against baseline approaches using historical population records.

Main Results:

• The novel technique tripled the match rate compared to baseline methods.
• Significant improvements in match accuracy were observed.
• The method demonstrated superior performance in linking fragmented or erroneous historical records.

Conclusions:

• The developed pre-processing technique offers a substantial improvement for historical Chinese record linkage.
• This approach has potential applications for other character-based languages like Japanese and various ethnic minority datasets.
• Language-specific data linkage strategies are promising for enhancing historical data usability across diverse populations.