Page 102 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 42

WU Wenna & BAO Xiulin / The architecture and data model of the National Thesauri Warehouse 101

semantic network can be viewed as an upper ontology and, to some extent, is useful in knowledge
inference. In contrast, the upper category system in STKOS is developed based on DDC. Thus, it is
a typical subject or discipline classification system. Its hierarchical relationship is not strict genus-
species relationship. Also, the semantic descriptions of its categories are not clearly and formally
expressed. Therefore, this category system cannot be used in knowledge inference. But it can
assemble concepts according to the subjects and disciplines. This is beneficial for classification-
subject integration during information process.
According to the integration modes as discussed above, the mapping center mode adopted by
HILT can be controlled easily. The services and applications of thesauri integration results can be
developed based on the central thesaurus. However, this mode claims higher demands on the central
thesaurus. Central thesaurus is expected to cover satellite thesauri across different disciplines and
cover their concepts of universality as well as specificity. The knowledge structure and descriptive
pattern of the central thesaurus need to be compatible with the satellite thesauri. Multi-thesauri inter-
mapping mode, as adopted in the KoMoHe project, presents higher complexity. It is difficult to
ensure the completeness of mapping. The development of mapping center is almost entirely random.
The development of services and applications of thesauri integration results lack main structure
to rely on. In comparison, the OCLC thesauri mapping project has characteristics of both above
modes. UMLS and STKOS adopted an integration mode of construction concept layer above the
source thesauri. Similar integration mode, called National Thesauri Warehouse, was proposed by Fu
Lansheng et al. in the 1990s (Fu, 1991; Zhu, 1991). This integration mode produces a unified concept
set that facilitates services and applications of integration results. Instead of having their own relation
network, the knowledge systems inherit relations from the source thesauri. Consequently, their
hierarchical structure can be overly complex and intertwined. Although it can be compensated by
back-tracing and utilizing relations in source thesauri, considering the existence of a large number of
source thesauri and complexly tangled relation network, it can possibly bring heavy burden on users.

2 NTW integration mode and architecture

Through investigation and comparative analysis of the integration modes of knowledge
organization systems in China and abroad, and considering the practical situation of thesaurus
construction in China, NTW adopted Chinese Thesaurus (CT) as the central thesaurus in mapping.
Two reasons are considered for making the decision and as shown bellow.
(1) CT is well compatible with other thesauri
In 1975, the CT project was initiated as an auxiliary project of the “748 Engineering Project”.
Based on the successful development of CT, a series of national thesaurus standards were
developed gradually. They include GB 3860-83 Document Indexing Standard, GB 13190-
91 Chinese Thesaurus Construction Standard, GB/T 15417-94 Multi-language Thesaurus
Construction Standard. The majority of Chinese thesauri were constructed after CT, and to some