语料库语料库研究方法概述.pptVIP

选题、设计与方法Put it altogether 李文中 中国外语教育研究中心 2012 语料库不是人学的， 正则表达式不是女人学的。 Corpus-driven is basically corpus based. Any corpus-based research is necessarily driven by corpus data. 目标：通过语料库分析和研究： 验证假设、直觉 获得新发现 建立新的假设 构建新的理论 验证已有的发现 解决难题 创新： 基于语料库方法是一种验证程序 语料库驱动方法是一种发现程序 理据：任何感知都是推断 Any perception is but inferencing. 基本步骤： 确定题目 提出问题 确定总体和样本 选择工具 处理数据 描述结果：分类、总结特征（description） 解释结果：观察、描述、解释（explanation） 解读结果：意义、价值、应用（interpretation） Identifying a problem Some thing or phenomenon: out of expectation Incongruent Need a solution puzzling Reading to be better informed What has been done as contribution What has been left undone What has been done wrong Never count someone else’s money. Formulating research questions Naming: what is… Classificatory: How are they interrelated (patterned)? Explanatory: to what extent do they co-occur? Predictive: What will happen if…? Never ask a question to which you already know the answer；never ask how to question Finding a method Population Sample Sampling IF P ? S S ? R R ? I THEN I ? P Descriptive research single text text vs. text people vs. text Research questions How many different word forms are used in the text? How many running words are used? What is their distribution? To what extent can the level of difficulty of the text be computed on the basis of the graded wordlists? How many different word classes are used? What is the number of each word class? Method To answer RQ 1, generate a wordlist of the given text and observe: The number of types The number of tokens the type/token ratio (TTR) If the text is very large, standardize the TTR the types and their frequency cumulative percentage To answer RQ 2, compute the wordlist against a batch of graded wordlists, and observe: How many types on Level 1, 2, and 3 lists are used in the text? And what is their percentage? What about their tokens? How many types that are not on any list are used in the text? Summarize their features. To answer RQ 3, retrieve each