索引典的中英詞彙語義對應之研究：以中國藝術領域為例

A Study of Mapping Chinese-English Lexical Semantics for a Thesaurus in Chinese Art

陳淑君(Shu-Jiun Chen)

指導教授 ： 陳雪華

國立臺灣大學/文學院/圖書資訊學研究所/博士(2012年)

https://doi.org/10.6342/NTU.2012.01868

摘要

知識組織系統的語義互通性，因為攸關使用者的跨資源檢索效能，是數位圖書館的重要研究議題之一。語義互通性是讓不同系統與服務能夠溝通資料、資訊與知識之際，並確保其正確性及保留其原來意涵。然而任何涉及多語言的知識組織系統，基於語言與文化的差異性，特別是需要整合不同文化觀點時，會產生相當大程度的複雜性。本研究希望從建構一個提供英文使用者檢索中華文化藝術資源的雙語索引典之角度，藉由分析中英文術語之語義對應與概念結構對應問題，探索中國藝術領域的重要詞彙在目前西洋藝術領域為基礎的索引典之語義互通性與方法。

本研究以台灣的國立故宮博物院器物處與書畫處總計2,100個中文控制詞彙（NPM），及美國的蓋堤研究中心「藝術與建築索引典」（AAT）共計35,314個英文概念為研究對象，以內容分析法進行中英文詞彙語義對應研究。主要研究問題有三：（一）中國藝術領域的控制詞彙與西洋藝術為基礎的索引典之間，到達什麼程度的語義互通性?（二）中國藝術領域的中英雙語之概念結構，有什麼樣的互通性?（三）對於無法精準對應的中文詞彙，如何解決以讓英文使用者理解其意義? 根據研究資料分析歸納，本研究主要發現有三，包括：

(一) 中國藝術領域詞彙對應到西方通用的全球藝術領域索引典時，其中三分之一具有「精準等同」對應關係，其它依序包括「狹廣階層」、「聯集等同」、「近似等同」、「交集等同」及「廣狹階層」等對應關係。其中，五分之三的詞彙為「狹廣階層」對應，因此若直接採用AAT作為中國藝術領域藏品的索引或檢索功能，系統的索引專指度將明顯不足，以詞彙量而言，尤以「代理者層面」的生物詞彙、「物件層面」的局部形制詞彙，及「物理特質層面」的紋飾詞彙最為缺乏。

(二) NPM對應至AAT的概念結構異同性之模式有四：第一種，有專門一類可以對應，且詞彙相當完整（如：中國考古學文化）；第二種，有專門一類可以對應，但詞彙不完整（如：中國陶瓷窯系）；第三種，無專門一類可以對應，且詞彙不完整（如：中國繪畫技法）；第四種，無專門一類可以對應，且無詞彙可以對應（如：中國傳說動物）。此現象意謂AAT目前已收錄部分專屬中國藝術領域的詞彙及結構（第一種），但仍有許多詞彙收錄尚未完整（第二種），或者還未賦予專門子類以集中相關詞彙（第三種），甚至，缺乏可以對應的概念結構與完整詞彙（第四種）。此外，本研究發現，NPM的部分概念結構與AAT不同，可能產生詞彙對應過程，無法完整保留其中一套詞彙的概念結構。

(三) 根據前述研究結果，針對無法精準對應的中文詞彙與概念結構，本研究提出一套實作方法架構，以讓英文使用者理解其意義，包括：「詞彙蒐集」、「文獻保證」、「範圍註增補」、「語義關係」、「結構調合」、「概念新增」，及「專家審訂」等七個模組，視詞彙不同的對應狀況，而有不同的模組組合，以達成最佳化的對應品質及語義互通性。此外，本研究針對異質概念結構，提出四種調合模式，以滿足雙方結構的完整性，或取得兩套詞彙架構的最佳化調整結果。

本文的研究結果將可貢獻於多語知識組織系統的語義互通性研究，特別是處理雙語索引典中，無法精準對應到英文的中文詞彙，深入而有系統地了解核心的問題，並嘗試提出解決方法，以滿足英文使用者檢索中文藝術資源之需求。

關鍵字

多語索引典 ； 知識組織系統 ； 控制詞彙 ； 語義互通 ； 中國藝術領域

並列摘要

The semantic interoperability between knowledge organization systems has significant bearing on the efficiency of cross-resource retrieval for users, which is an important issue for digital library research. Semantic interoperability allows different systems and services to communicate data, information and knowledge, ensuring its accuracy and retaining their original meaning. However, any multilingual knowledge organization system, based on language and cultural differences-- especially with the need to integrate different cultural perspectives--will generate a considerable level of complexity. For the purpose of building a bilingual thesaurus enabling English users to access Chinese culture and art resources, this study aims to explore the semantic interoperability and methodology for important Chinese art-related terms in present Western art-based thesauri by analyzing the semantic mapping problems of Chinese and English terminology.

This study takes a total of 2100 Chinese controlled vocabulary from the National Palace Museum (NPM) in Taiwan, and a total of 35,314 English concepts of Art & Architecture Thesaurus (AAT) developed by Getty Research Institute in the U.S., and uses them as research objects; adopting content analysis to conduct a semantic mapping between Chinese and English terms. There are three main research questions: (1) To what degree of semantic interoperability can be achieved between Chinese art-related controlled vocabulary and Western art-based thesauri? (2) What patterns of interoperability are involved in the field of Chinese art in the bilingual conceptual structure? (3) For those Chinese terms that cannot be exactly matched, how to make English users understand their meanings? According to the analysis of research materials, the main findings of the study include:

(1) When mapping Chinese art-related terms to an international art thesaurus commonly used in the West, one-third of them can be mapped with “exact equivalence”, while other mapping relationships encompass “hierarchical mapping (narrow to broader)”, “cumulative compound equivalence”, “inexact simple equivalence”, “intersecting compound equivalence” and “hierarchical mapping (broader to narrow)”. Among them, three-fifths of the terms are mapped with “hierarchical mapping (narrow to broader)”, so it is clear that directly using AAT to index or retrieve Chinese art-related collections will lead to insufficient indexing specificity of the system. In terms of the quantity of vocabulary, the biological terms in “Agents Facet”, the terms about parts of objects in “Objects Facet”, and the motif terms in “Physical Attributes Facet” are most lacking.

(2) There are 4 different structural models that NPM corresponds to, regarding the concept of AAT: The first pattern is a specialized group that can be mapped, and the vocabulary is fairly complete (ex: Chinese Archaeology Culture); The second pattern is a specialized group that can be mapped but the vocabulary is incomplete (ex: Chinese ceramic styles); The third pattern is a specialized group that cannot be mapped, and the vocabulary is incomplete (ex: Chinese painting techniques); The fourth pattern is a specialized group that cannot be mapped but the vocabulary is complete (ex: Chinese mythical creatures). This result means that AAT has already embodied a portion of vocabularies and structures that exclusively belong to Chinese arts, (the first group), but there are still many included vocabularies that are incomplete (the second group) or have not been given a special subclass to focus on relevant vocabulary (the third group), even with the lack of the concept of structure, this group can be mapped with a complete vocabulary (the fourth group). In addition, the study found that the NPM part of the conceptual structure with AAT may be unable to keep intact the conceptual structure of one particular set of vocabulary during the process of producing corresponding vocabulary.

(3) According to the abovementioned research finding, the study proposes a set of methodological framework for Chinese terms and conceptual structures that cannot be exactly matched, in order to let English users understand their meanings, which includes seven modules: term collection, literary warrant, scope note supplement, semantic relationships addition, structure adjustment, concept creation, and expert examination. There will be different combinations of modules based on different mapping results of terms, so as to reach the optimum mapping quality and semantic interoperability. In addition, this study proposes four synthesizing methods for heterogeneous conceptual structures to meet the structural integrity between the two sides, or to obtain the best adjustment results in the two sets of vocabulary structure.

The findings of this study will help contribute to the research on the semantic interoperability between multilingual knowledge organization systems, particularly in dealing with Chinese terms that cannot be mapped to English in multilingual thesauri. The study probes deep into the core problems in a systematic way, trying to find a solution in order to satisfy the need of English users for accessing Chinese art-related resources.

並列關鍵字

Multilingual Thesaurus ； Knowledge Organization Systems ； Controlled Vocabularies ； Semantic Interoperability ； Chinese Art

參考文獻

謝明良（2005）。臺灣海域發現的越窯系青瓷及相關問題。臺灣史研究，12(1)，115-163。

陳光華（2000）。數位圖書館中權威控制系統的設計。政治大學圖書與資訊學刊，34，51-71。

吳子淦（2001）。花虻：作物授粉及生物防治的好幫手。農業試驗所技術服務，NO.

Spärck Jones, K. (2007) Semantic primitives: the tip of the iceberg, In Words and intel-ligence: Part II: Essays in Honour of Yorick Wilks, (Eds. K. Ahmad, C. Brewster and M. Stevenson), Springer: Berlin.

Aitchison, J, Gilchrist, A. and Bawden, D. (2000). Thesaurus construction and use: A practical manual. (4th ed). London: ASLIB/IMI.