| Abstract: | 在這個台灣華語分類詞的三年期研究計畫中，我們企圖達成三項目標。建立關於此議題的詳細書目為我們的第一個目標。其次，我們計畫將中央研究院現代漢語平衡語料庫、中文十億詞語料庫Chinese Gigaword（經由Sketch Engine）、台灣地區的Google 搜尋、《國語日報量詞典》以及《實用量詞詞典》等來源之分類詞語料做統合整理，並建構出台灣華語分類詞的專屬次語料庫。而最重要的目標則是提出對分類詞與量詞的合理形式分析以及語意認知的分析；前者我們將在詞彙功能語法（LFG）的理論架構下，提出形式句法的分析，後者我們將利用認知語法中的框架（Frame）及顯影（Profile）的概念做出認知語意上的分析。（此為2 年期整合型計劃「台灣華語：書目、語料庫與教學參考語法」下之子計畫之一。此子計畫為3 年期的原因在於第3 年的認知語意的分析。前2 年的書目、語料庫及形式語法分析則與整合型計劃及其他子計畫的步驟完全相符。）基於分類詞是否有別於量詞的論戰，在次語料庫的建構及分析之前，我們必須先確立分類詞之所以有別於量詞之具體特徵。在語意方面，我們首次運用亞里斯多德對於必然性質/偶然性質之區分，以及康德對於分析命題/綜合命題的區分來精準描繪分類詞/量詞之區分。在句法方面，支持區分分類詞/量詞的學者曾提出兩個測試：形容詞修飾及「- 的」的插入；然而此二測試已被反對區分的學者證實為不可靠。因此，我們對此二測試做了更深入的檢視，修正為精準無誤的句法測試，從而確認區分分類詞/量詞的真實性。文獻上對於分類詞的數量以及所列清單均存有極大的差異。因此，我們將依據上述的語意及句法特徵，建立台灣華語分類詞的清單。為了取樣的精準並減少研究者語法判斷上的爭議，我們也將以18-25 歲之台灣華語母語人士為受試者，進行一連串心理語言學上的測試，作為選定分類詞的依據。同時，我們也將建立並參考台灣華語分類詞的專屬次語料庫，以及關於此議題的相關文獻。句法分析上，我們將在詞彙功能語法的架構下提供一非以移位為基礎的分析，來解釋以下的事實：分類詞與量詞在句法上佔據不同的位置，且兩者並非互斥關係，[數詞- 分類詞] 與 [數詞-量詞]各自形成詞組，量詞且在結構上高於分類詞，而分類詞與量詞均可以形容詞修飾。我們也將論證，[名詞-分類詞]的組合（如：米粒、馬匹、花朵、槍枝）並非生成於句法，而是詞彙。此外，我們也期盼此一句法分析能反映出下述認知語意分析的內涵，例如：不同於實詞身分之量詞，分類詞僅為一功能詞，因此對於整個片語並不貢獻任何實質意義。句法分析以外，我們亦將檢視名詞所選擇搭配之分類詞隱藏於語意及認知概念下的動機。首先，以Fillmore (1982)的框架語意學為基礎，建構出名詞框架的概念並用來指涉名詞的語意特徵結構，進而結合Langacker (1987) 的顯影原理作為選擇分類詞的認知機制。同時，我們將一併檢視不同類別的分類詞如何相互影響、相互競爭。以台灣華語的討論為基礎，我們企圖以名詞框架及顯影原理的架構作為分析所有分類詞語言的一套可靠工具。此一架構最先由主持人於2008 政大漢語語言學課堂上提出，Hsieh（2009）在主持人所指導的碩士論文中，做出進一步的研究。在此基礎上，我們將對台灣華語分類詞做出全面的認知語意分析。本計劃將分三階段進行，每階段大約為期12 個月。第一階段將依序建構文獻書目、分類詞清單以及次語料庫，全數資料將上傳於計劃網頁上開放取用。第二階段為形式句法分析，第三階段則著重於建立分類詞的認知語意分析。主持人將主導本計劃的每一面向，但Co-PI 萬依萍將共同負責心理語言學測試的設計及分析，而Simon Smith（Chinese Gigaword 及Sketch Engine 的研發者之一）則負責建構分類詞專屬的次語料庫。
In this 3-year study of classifiers in Taiwan Mandarin (TM), we aspire to achieve three goals. First, we aim to compile a comprehensive bibliography for this subject matter. Second, a special purpose subcorpus of classifiers in TM will be constructed, drawing from Sinica Corpus, Chinese Gigaword (via Sketch Engine), Google searches in the Taiwan domain, as well as《國語日報量詞典》(Mandarin Daily News Dictionary of Chinese Classifiers) and《實用量詞詞典》(A Practical Dictionary of Classifiers). More importantly, we aim to provide both a formal syntactic analysis, within the Lexical-Functional Grammar (LFG), and a comprehensive functional account using the frame/profile approach in cognitive grammar. Given the debate whether there is a distinction between classifiers (C) and measure words (M), it is crucial that we first establish concrete criteria to distinguish C/M. This step is logically prior to the subcorpus construction and analyses. For semantics, we shall for the first time employ the Aristotelian distinction between essential and accidental properties as well as the Kantian distinction between analytic and synthetic propositions to characterize the C/M distinction. For syntax, there are two tests, adjectival modification and de-insertion, proposed by proponents for the C/M distinction but deemed to be unreliable, if not unworkable, later by opponents. We shall examine the two tests much more closely than previous studies and revise them and come up with two refined, reliable, and revealing syntactic tests. We shall thus demonstrate unequivocally that the C/M distinction is real. To rectify the drastic disparity in the inventories of classifiers proposed in the literature, we will first restrict our inventory to TM only and base the selection on the semantic and syntactic criteria proposed above. To further increase the prudence of this inventory and to resolve controversies over grammaticality judgments, we will conduct psycholinguistic tests on young adult (age 18-25) native speakers of TM. Corpora will also be consulted and a special purpose subcorpus will be constructed along the way. Meanwhile, a comprehensive bibliography on this subject will also be compiled. In terms of syntactic analysis, we intend to offer a formal non-movement-based account in the lexicalist LFG framework that accounts for the following facts: C and M occupy different syntactic slots and are thus not mutually exclusive, [Num-C] and [Num-M] form a constituent respectively, M is structurally more prominent than C, both C and M can receive adjectival modification, and the sporadic N-C forms, e.g., 米粒 mi3li4 ‘rice grain’ and 馬匹 ma3pi1 ‘horses’, are lexical, not syntactic. Furthermore, the account will be able to reveal the insights from the functionalist account proposed below, e.g., unlike M, which is substantive, C is a functional category and does add to the meaning of the phrase. We shall also look beyond syntax and examine the semantic and cognitive motivation behind the choice of Cs for a noun. We will establish the concept of N-frame, based on Fillmore’s (1982) frame semantics for verbs, to refer to the feature-structure of a noun, and then integrate Langacker’s (1987) principle of profiling as the cognitive mechanism for classifier selection. We will also examine how different classes of classifiers interact and compete. Based on the discussions of Taiwan Mandarin, we aspire to establish the N-frame/Profiling framework as an insightful analytical tool for all classifier languages, a framework first proposed by the PI in a lecture in 2008 and explored in Hsieh (2009), an MA thesis the PI supervised. The project will proceed in three phases, each roughly corresponds to a 12-month period. We shall establish the bibliography, the classifier inventory in TM, and the subcorpus in the first phase and all will be made open access on the project web site. We then proceed to the formal LFG analysis in the second year, followed by the third phase of constructing a comprehensive N-frame/Profiling-based system for TM. While the PI is to oversee all aspects of the project, co-PI I-Ping Wan will take charge of the psycholinguistic tests and data analysis, and co-PI Simon Smith, an original developer of Sketch Engine, will supervise the construction of the special purpose subcorpus. |
