跨語言空殼名詞平行語料庫建置

The Construction of the NCCU Shell Nouns Cross-Language Parallel Corpus (NCCU-SNPC)

「空殼名詞」－即意義相對空泛的名詞─其出現通常泛指非特定的想法、問題等名詞，必須透過文章內容得知其指涉的意義為何。跨語言結構資料庫為因應不同目的而建構的資料庫與日俱增，然現存的資料庫中，並無提供空殼名詞的資源。本文章提供建置「空殼名詞跨語言結構資料庫」之內容，其中包含五種語言（中文、日文、捷克文、德文以及越南語）的空殼名詞跨語言對應。執行流程：1.收集語料；2.空殼名詞翻譯：以Schmid（1999；2000）的研究為本，將英語空殼名詞翻譯為中文、日文、捷克文、德文以及越南語；3.五種語言空殼名詞對照；4.語料庫介面建立；5.建置「空殼名詞跨語言結構資料庫」完成。語言學研究需要共享的語言資料，如各類詞表。這些詞表的共享激勵了相關研究能量，其共通點是可以利用電腦程式處理，並可成為人工智能或數位學習程式的應用資源。本研究所提供的提供人工標註翻譯及檢索平台可作為未來更多空殼名詞或跨語種研究的資料，應用於人工智慧或數位學習資源。

The concept of a shell noun can be understood as a noun having a bare meaning to express an uncertain concept. This involves words such as idea, problem, issue, and some other general nouns. Normally, most shell nouns are highly contextualized. Their meanings are largely based on the context where they appear. Their existence in other languages, however, is overlooked. At present, cross-language structured databases have been constructed to meet various purposes, and they continue to grow. Yet, among current corpora, we rarely see a multiple-language translation of shell nouns. For this purpose, it is worthwhile to conduct a database with example sentences that can compare their uses in various languages. The aim of this project is to create one such database–the NCCU Shell Nouns Cross-Language Parallel Corpus (NCCU-SNPC), which includes data from Chinese, Japanese, Czech, German, and Vietnamese. These are the steps: (a) We first collected related language data from Schmid (1999, 2000). (b) The language data was translated into the target languages, based on the list and sentences we collected. (c) The mapping of different languages was carried out. (d) A platform containing the five languages with searchable function was created. (e) The final step was to finalize the construction of the NCCU-SNPC and to check for consistency. Linguistic studies require shared data of different languages, and it has become a trend to share data to inspire related research. Using our data, more research can be carried out so as to provide human annotated data for applications in A.I. or digital learning resources.