A Data-driven Approach Based On Keywords Analysis In Discourse Analysis

A data-driven approach in corpus linguistics allows a specific research question to be identified in the course of a research project rather than at the outset (Rayson 2002). Keywords analysis is one central research area in corpus linguistics, which provides a useful way to characterise a text or genre (Scott 1996). This thesis makes an attempt to unite these two concepts and suggests practical procedures for corpus linguists to adopt a data-driven approach based on keywords analysis in discourse analysis.This thesis first reviews the literatures on corpus linguistics, three research approaches in corpus linguistics (corpus-based, corpus-driven and data-driven), and three main methods in corpus linguistics (keywords analysis, collocation and concordance).Based on the previous analysis, the author proposes a four-step data-driven approach based on keywords analysis, that is, "build - retrieve - question - interpret", and subsequently tests it in a case study of China's economy as reported by western scholars.Preliminary research indicates that a data-driven approach based on keywords analysis is capable of directing us to find unexpected linguistic phenomena that we would not otherwise have considered for study.Problems of the data-driven approach based on keywords analysis are nevertheless experienced in the process of conducting this research. Possible improvements to be made in future researches are suggested accordingly.The implications of this study are directed at discourse analysis, language learning/teaching and information extraction.