從自然語言處理觀點探索大學部EFL學生於英中逐步口譯使用言談標記之情形

Adopting Natural Language Processing to Explore the Use of Discourse Markers by EFL Undergraduates in English-to-Chinese Consecutive Interpreting

自上世紀七十年代開始，就有許多如何在自然語言情境中使用言談標記以利言談管理的研究出現，而這些研究也給予言談標記不同的名稱，諸如句子連接詞 (Halliday & Hansan, 1976)、話語連接詞 (Blakemore, 1987)、話語運作因子 (Redeker, 1990, 1991)、談話口語詞 (Hansan, 1998) 等。除了在自然語言情境中使用言談標記的研究之外，近年也有些研究把目光投向了言談標記在同步口譯中的使用情形，這些研究指出，在同步口譯中使用言談標記可被視為一種策略，用以提升譯語的流暢度以及聽者對訊息的理解力 (Zheng, 2013; Tian, 2015; Wang, 2016)。相較之下，關於逐步口譯中如何使用言談標記的研究卻相對有限，故，本研究以此為出發點，使用了11位大學部EFL學生在英譯中模擬逐步口譯考試時的音檔為研究資料來源，剖析他們如何使用言談標記，並透過中研院的線上分詞系統探討這些言談標記提供何種語用功能。本研究發現，「這」跟「來」是最常被加入至譯語中的言談標記，其次為「然後」與「這些」。本研究也發現學生們在譯語中一共加入了17個言談標記，並由線上分詞系統解構後發現，這17個言談標記的語用功能取決於其出現的語境為何，也就是說，即便是同一個言談標記，其語用功能並非都是維持一致，端看語境而定。

 

Discourse markers have been researched extensively in spontaneous speech since the 1970s, with quite a few different names given to manifest their functions in discourse management, such as sentence connectives (Halliday & Hansan, 1976), discourse connectives (Blakemore, 1987), discourse operators (Redeker, 1990, 1991), and discourse particles (Hansan, 1998). Apart from spontaneous speech, some recent studies have also shed light on the use of discourse markers in simultaneous interpreting as a coping strategy to enhance understanding and fluency (Zheng, 2013; Tian, 2015; Wang, 2016). Nonetheless, more research is needed in the scenario of both consecutive interpreting and EFL undergraduate classroom. Therefore, this study exploited voice recordings in a mock mid-term exam from 11 EFL undergraduates to investigate how they used discourse markers over the course of English-to-Chinese consecutive interpreting to explore the discourse functions of the markers from the point of natural language processing with the help of an online Chinese parser. The study found that zhe (this) and lai (come) were the two most frequently added discourse markers, followed by ranhou (then) and zhexie (these). A total of 17 discourse markers were identified as additions to the rendered texts. All the added discourse markers assumed individual discourse functions that were context-bound as identified by the parser.