張國恩宋曜廷張道行陳于佳Yu-Chia Chen2019-08-292015-07-312019-08-292012http://etds.lib.ntnu.edu.tw/cgi-bin/gs32/gsweb.cgi?o=dstdcdr&s=id=%22GN0699080073%22.&%22.id.&http://rportal.lib.ntnu.edu.tw:80/handle/20.500.12235/93084由於全球學習華語熱潮的興起，學習華語的人數年年增加。不論是母語學習，或者是第二語言學習，閱讀在語言學習中都扮演了重要的角色。但如何選擇適當難易度的文本是在閱讀過程中常面臨到的問題。教師在教學時必須選擇符合學習者程度的教材，來提高華語文教學的效率。 本研究以《新版實用視聽華語》、《遠東生活華語》、《新實用漢語課本》、《中文聽說讀寫》、《讀報學華語》、《實用商業會話》等六套常用華語教材為例，結合特徵選取方法與支援向量機建立預測模型預測文本CEFR等級，並探討不同特徵組合所造成結果的差異。 實驗結果顯示，不需使用全部指標，只須採用較為重要的指標組合即可達到最佳預測正確率，約為85.47%。期盼本研究所建立的可讀性算則不但可提供華語教師能更系統化、循序漸進的教學，學生也可透過此預測模型選擇符合自身程度的課文學習，減少摸索的時間，來達到有效提升學習成效的結果。In recent years, the number of people in the world learning Chinese is growing rapidly. Reading plays an important role in language learning. But how to select reading text which is suitable for learners is one of problem in reading. Teachers have to choose reading materials at right reading level for learners to improve efficiency. Readability assessment is a method to quantify reading difficulty for learners. In this study we combined support vector machine with feature selection methods to construct a model to predict the CEFR level of the six most popular Chinese teaching materials : Practical Audio-Visual Chinese, Far East everyday Chinese, New Practical Chinese Reader, Integrated Chinese, Learning Chinese with Newspaper, and Practical Business Conversation. We also compared the predicting performance by different combination of features. The experimental results have shown the effectiveness of the feature selection method. Choosing important features can reach the best performance, the accuracy is about 85.47%. We hope our study can promote more effective teaching and learning in Chinese learning.可讀性特徵選取支援向量機ReadabilityFeature SelectionSupport Vector Machine