標題: 以框架語義為本之中文動詞詞彙語義研究:中文陳述性動詞之分析
A Frame-based Corpus Approach to Mandarin Verbal Semantics: The Case with Mandarin Statement Verbs
作者: 江亭儀
Ting-Yi Chiang
劉美君
Mei-Chun Liu
外國語文學系外國文學與語言學碩士班
關鍵字: 動詞詞彙語意研究;中文陳述性動詞;框架語義;近義詞分析;Verbal semantics;Chinese Statement verbs;Frame Semantics;Near-synonymy
公開日期: 2005
摘要: 動詞詞彙語意研究近年來一直是語言學重要的研究議題之一,動詞詞彙語意研究與句法和語意之間的互動有著緊密的關係。此篇研究採用Liu et al. (2004) 提出的多層級研究框架,此研究架構以框架語義為本 (Fillmore and Atkins 1992),總共包含5個層級:領域-框架-次框架-近義詞詞組-詞彙,希望可以對中文動詞語義研究提供完整的分析模式。使用此研究架構,即可以對中文陳述性動詞中包含各種不同動詞類型,提供一完整有系統的分析。在中文裡,陳述性動詞包含非常多詞彙,例如:說、講話、表示、說明、建議等…陳述性動詞是溝通類動詞裡非常重要且也是主要的成員之一。中文陳述性動詞典型的句法表現為:動詞前的主語為說話者,動詞後帶的受詞為說話者說出的內容或訊息,以下例子可以說明此句法表現: (1) 她[Speaker] 說/表示/解釋/建議/抱怨/透露/提醒 那麼你自由自在的飛吧 [Message] 雖然這些動詞都展現類似的句法形式,但他們卻表達非常不同的溝通事件,於是一項基本的議題成為本研究的主要重點:這些動詞彼此之間有什麼不同?更具體的說,這些動詞在語義上有何區分,句法和語義之間存在著什麼樣的關連關係? 採用5層級以框架理論為基礎的研究架構 (Liu et al. 2004),中文陳述性動詞在定義上是屬於陳述框架,而這個框架屬於一個更上層、更大的知識本體—溝通領域,陳述性動詞在認知上皆牽涉到一項重要的概念:表達單向性的訊息包裝與傳達之溝通事件,屬於這個frame且用來定義這些動詞的框架元素為說話者、訊息本身、主題、聽話者與溝通媒介。這麼多陳述性動詞可以依據框架元素之凸顯和隱藏與否與其在句法上之展現,將這些動詞作次分類,即區分次框架,同時也是在每個次框架裡,可以再進一步區分近義詞詞組,區分中文陳述性動詞的標準總共有十一項,羅列如下:1)名物化的比例 2)動詞詞彙化包含訊息之句法表現 3)主題使用之不同 4)聽話者出現之差異 5)溝通媒介之使用 6)說話者角色之差異 7)訊息類型之不同 8)說話者和聽話者之間的關係 9)人為溝通事件的主題 10)陳述動詞與其他動詞合用的V+V句法表現 11)與時態標記共現的差異。根據以上這十一項區分標準,可以將58個中文陳述性動詞進一步區分為九組subframes: 1) subframe 1- 說類; 2) subframe 2- 說話類; 3) subframe 3- 表示類; 4) subframe 4- 說明類; 5) subframe 5- 建議類; 6) subframe 6- 透露類; 7) subframe 7- 抱怨類; 8) subframe 8- 承認類; 9) subframe 9- 報告類。 採用這樣的研究架構,對中文陳述性動詞的研究可以大量而且深入,同時,也可以對動詞的語義表徵作比較有系統且精確的分析,此研究顯示動詞語義研究應該是具有階層性的,而且跨語言的動詞語義研究也可以利用此研究框架來進行,此研究最終是希望能對中文與跨語言的動詞詞彙語義研究有所貢獻。
Verbal semantics has been an important issue in linguistic research that deals with the interaction between syntax and semantics. This research adopts the frame-based (Fillmore and Atkins 1992) classification hierarchy (Domain > Frame > Subframe > Near-synonym Set > Lemma) proposed by Liu et al (2004) and provides a unified solution to the study of Mandarin verbal semantics.With the case of Chinese statement verbs, the ‘five-layered hierarchical model’ is shown to be effective and comprehensive in exploring the granularity of verb types. In Mandarin, verbs of statement which include a wide variety of verbs, e.g., shuo說 ‘say’, jianghua 講話 ‘talk/speak’, biaoshi表示 ‘express’, shuoming說明 ‘explain’, jianyi 建議 ‘suggest’, and other words form a major group in the Communication Domain. As pointed out by Liu (2003), verbs, including verbs of statement in the Communication Domain encode the most fundamental aspect of human activities. Prototypically, verbs of statement tend to take a Speaker as the subject and a Message the object, as illustrated below: (1) 她[speaker subject] 說/表示/解釋/建議/抱怨/透露/提醒 那麼你自由自在的飛吧 [message object] Although these verbs share the basic surface pattern, they encode very different communicative events. One fundamental question in this study is: how do these verbs differ from each other? More specifically, what are the distinctions in their lexical semantics and syntax-to-semantics correlates? The analysis of the present study is corpus-based. All the data and statistics of this paper were mainly based on “Academia Sinica Balanced Corpus of Modern Mandarin Chinese (Sinica Corpus)”. The distributional tendencies rather than grammaticality were taken as the important evidence for linguistic analysis. Following the five-layered frame-based framework (Liu et al. 2004), a large set of Chinese statement verbs is first characterized with a cognitive schema of the Statement frame and linked to the upper semantic ontology, the Communication domain. Verbs in the Statement frame all involve “unidirectional message-transferring” and can be defined by a set of core frame elements: Speaker, Message, Topic, Addressee, Means. Corpus observations of the major syntactical realizations of the frame elements (Basic Patterns) and collocational associations are then analyzed to categorize verbs into several subtypes, i.e., subframes. Under the frame level, syntactic foregrounding and backgrounding of certain core frame elements and their syntactic manifestations were taken as the basis for further categorizing verbs into nine subframes. It is also at the subframe level that all the near-synonym sets are anchored. By first investigating the high-frequent verbs in detail, eleven criteria for further dividing the subframes were found: 1) Frequency of nominalization 2) Incorporated Message 3) Distributional contrasts of Topic 4) Distributional contrasts of Addressee 5) Overt marking of Medium_Means 6) Variation of Speaker role 7) Variations of Message types 8) Relation between Speaker and Addressee 9) Topic as Human 10) V+V pattern 11) Collocational variations of aspectual modifications. Based on the above eleven criteria, fifty-eight verbs of statement are further categorized into nine subframes, each with a unique set of highlighted frame elements, basic patterns, collocational associations, and semantic features. The nine subframes of the statement frame are: 1) subframe 1- Say subframe; 2) subframe 2- Speak subframe; 3) subframe 3- Express; 4) subframe 4- Explain; 5) subframe 5- Suggest; 6) subframe 6- Reveal; 7) subframe 7- Complain; 8) subframe 8- Admit; 9) subframe 9- Report. Finally, the syntactically-motivated manifestations will be attributed to their semantic attributes and will be represented in Module-Attribute Representation of Verbal Semantics (MARVS) proposed by (Huang et al 2000). It is shown that verbal semantics should be represented and linked to a well-structured hierarchical framework. Ultimately, this thesis shows that this model can not only capture a fine-grained semantic representation (much as Pustejovsky (1995) has claimed—the semantic well-formedness) in Mandarin but also helps to explore verbal semantics in a cross-linguistic perspective.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009245516
http://hdl.handle.net/11536/77408
Appears in Collections:Thesis


Files in This Item:

  1. 551601.pdf
  2. 551602.pdf
  3. 551603.pdf
  4. 551604.pdf
  5. 551605.pdf
  6. 551606.pdf
  7. 551607.pdf
  8. 551608.pdf

If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.