Abstract
Institutional repositories (IRs) primarily rely on self-archiving, which often leads to inconsistent metadata, particularly regarding author names. This study addresses the challenges of information retrieval caused by name fragmentation and homonymy. By developing a "Name Access Point Control System" (NA_DSpace), the researchers moved away from traditional rigid authority control toward a flexible grouping method. This system utilizes all name variants found in source materials as access points and employs subject-area context to distinguish between authors with identical names. Implementation results demonstrate significantly improved search functionality and metadata consistency.
The Challenge of Self-Archiving and Name Fragmentation
Institutional Repositories (IRs) like DSpace are essential for the open dissemination of academic research. However, because they rely on "self-archiving"—where authors input their own metadata—they frequently suffer from data quality issues. Unlike traditional libraries managed by professional catalogers, IRs often lack standardized "authority control."
This decentralization leads to two primary issues:
- Name Fragmentation: A single researcher's name may be recorded in multiple formats (e.g., "Kim, Seong-Cheol," "Kim, Seongcheol," or "김성철").
- Homonym Confusion: Different researchers sharing the same name are often indistinguishable in search results.
Traditional library authority control, which designates one "correct" name form, is often too manual and rigid for the dynamic environment of an IR.
Methodology: The NA_DSpace System
To resolve these issues, the authors developed NA_DSpace, a modified version of the open-source DSpace platform. The system introduces several innovations:
1. From Authority to Access Points
Rather than enforcing a single "authorized" name, the system treats all variations as equal "access points." These variants are grouped under a unique identifier, ensuring that a search for any version of a name retrieves all associated works.
2. MADS Integration
The system utilizes the Metadata Authority Description Schema (MADS), an XML-based format. This allows for better compatibility with web environments compared to traditional MARC-based records.
3. Subject-Based Identification
To distinguish between homonymous authors, the system moves beyond birth years—which are often unavailable or private. Instead, it leverages the field of activity (subject) and work titles extracted from bibliographic data. For example, a user might see "Bae, Zong-Tae [Technology]" to distinguish him from another individual with the same name in a different field.
4. Real-time Metadata Entry
An AJAX-based interface provides real-time suggestions during the upload process. This "auto-complete" feature displays the researcher's name alongside their subject area, helping authors select the correct identity and preventing the creation of duplicate entries.
Key Insights and Results
The study utilized real-world data from Seoul National University (S-Space), KAIST (KOASAS), and Inha University to test the system.
- The "No-Authority" Philosophy: By acknowledging that name variations are natural and providing a system to link them, the researchers improved retrieval without requiring users to change their naming habits.
- Contextual Clarity: Using academic departments or subjects is more intuitive for users than birth years for identifying researchers.
- Efficiency: The auto-complete feature reduced the time required for metadata input while maintaining high data quality.
Limitations and Future Directions
Despite its success, the authors identified areas for further development:
- The "Student Gap": Many IR contributors are graduate students who lack existing records in library authority files. A more robust "onboarding" process for new researchers is needed.
- Maintenance: While highly automated, human oversight remains necessary to verify complex name groupings and ensure accurate subject classification.
- Global Integration: Future iterations should aim to link local IR systems with global identifiers like ORCID or ISNI to maintain name control across institutional boundaries.
Summary updated