以框架理論為本之漢語內在評價動詞詞彙語意分類研究—從「同意」的多義性出發

Abstract

本研究以Fillmore和Atkins(1992)所提出的「框架語意理論」(Frame semantics)及Liu和Chiang(2008)提出的「中文動詞語意網之架構」為本，以「同意」的多義性出發，探討漢語中的內在評價動詞(Internal Evaluation Verbs)如相信、懷疑、重視、輕視、支持、反對，並進行分類以及解決跨範疇詞彙歸屬議題。過去已有許多關於判斷(Judgment)以及評價(Evaluation)動詞的研究，Graeme Forbes (2006)討論過fear, distain, and worship；Longacre (1996)專注於disdain, derision, scornful, gossip, criticizing, maligning並將這些詞彙進一步分類為正面評價以及負面評價；Fillmore (1971)將評價動詞(Verbs of Judging)定義為一組隱涵人際關係互動以及責任歸屬的動詞，包含accuse, blame, criticize, credit, praise, scold, apologize, forgive, justify, excuse。 在中文方面，目前還尚未有任何關於評價動詞系統性的分類。本研究採取「中文動詞語意網」之四階層(主框架、首要框架、基本框架以及微框架)架構以及語料為本的分析將中文內在評價動詞做階層性的分類，而內在評價動詞為評價主框架下的其中一個首要框架。 在語料中，觀察發現「同意」有如下四種用法：

(1)Epistemic judgment: 我同意明天可能會下雨。 (2)Affective judgment: 我同意你的看法。 (3)Deontic judgment—Intention 我同意離開。 (4)Deontic judgment—Preference 我同意你離開。

藉由本研究，可發現「同意」的四種用法分屬於內在評價首要框架下的四個不同的基本框架。 內在評價動詞是評價主框架的首要框架之一。在內在評價首要框架下，共有四個基本框架及七個微框架。另外一個議題是關於動詞的「多元承繼」。例如內在評價動詞下的基本框架Certainty frame，同時承繼內在評價首要框架以及認知範疇(Hu 2007) Thought_creation state首要框架的特徵。 簡言之，本研究主要觀察漢語內在評價動詞在語料中所呈現的語意、語法特點，對漢語情緒動詞做進一步全面的分類，希望能為漢語內在評價評價動詞的研究提供清楚的分析；另外，本研究還將對框架語義理論中很少被提及的多元承繼的跨類現象做一個完整的解釋。

Based on Frame Semantics proposed by Fillmore and Atkins (1992) and the Framework of Mandarin VerbNet by Liu and Chiang (2008), this study attempts to explore Mandarin Internal Evaluation verbs like 相信xiangxin ‘believe’、懷疑huaiyi ‘doubt’、重視zhongshi ‘value’、輕視qingshi ‘belittle’、支持zhichi ‘support’、反對fandui ‘be opposed to’ from the perspective of the polysemy of 同意tongyi ‘agree’ and classify the verbs into a hierarchical structure as well as deal with cross-domain problems. In the literature, there are several researches on Judgment verbs and Evaluation verbs. For example, Graeme Forbes (2006) discusses the lexicons fear, distain, and worship; Longacre (1996) focuses on disdain, derision, scornful, gossip, criticizing, and maligning as well as divides those predicates into Verbs of positive evaluation and Verbs of negative evaluation; Fillmore (1971) defines Verbs of Judging as verbs which represent various types of interpersonal relationships involving judgments of worth and responsibility and those verbs include accuse, blame, criticize, credit, praise, scold, apologize, forgive, justify, and excuse. As for Mandarin, there is none systematic analysis for Judgment verbs so far. This study adopts a multi-layered hierarchical structure (Archiframe > Primary frame > Basic frame > Microframes) and corpus-based analysis to classify verbs in Internal Evaluation domain, which serves as a primary frame in Judgment domain. In the corpus, 同意tongyi ‘agree’ have four kinds of usage as the following shows:

(1)我同意明天可能會下雨。 wo tongyi mingtian keneng hui xiayu I agree tomorrow maybe will rain ‘I agree that it will be raining tomorrow.’ (2)我同意你的看法。 wo tongyi ni de kanfa I agree you DE opinion ‘I agree with you.’ (3)我同意離開。 wo tongyi likai I agree leave ‘I agree to leave.’ (4)我同意你離開。 wo tongyi ni likai I agree you leave ‘I agree that you leave.’

Through this study, it can be found that the four kinds of 同意tongyi ‘agree’ can be settled in four different basic frames under the same primary frame—Internal Evaluation. Internal Evaluation is the primary frame under Judgment Archiframe. For Internal Evaluation, four basic frames and seven microframes are included. Another issue is about the multiple inheritance of a frame. Certainty frame, one of the basic frames under Internal Evaluation, inherits the feature from both Internal Evaluation and Thought_creation state from Cognition domain (Hu 2007). From the figure, it can be seen Certainty frame inherits the feature of Cognition domain and Judgment domain at the same time. The lemmas in Certainty frame are not polysemous since the central idea of frame-based analysis is “one frame, one meaning”. Instead, they belong to different domains (Cognition and Judgment) because of the adoption of different perspectives from the observer. Through this study, in addition to providing a systematic analysis for Mandarin Internal Evaluation verbs, the issue about cross-domain lemmas can be solved.