心理語言學認為,閱讀是一種認知的過程,其過程包括將句子轉換成有意義的文字組塊。在本研究中,我們讓閱讀者藉著使用我們所架構的外顯閱讀系統,來讓隱藏在讀者的心中的語塊顯現出來。讀者可以利用滑鼠點擊在單字之間來切斷或是取消切斷一行單字,藉此來變成更多或更少行的單字,點擊單字或是每行單字前端的空格來進行縮排或是取消縮排,藉此建立階層式的語塊結構。 在這項研究中,我們成功的架構了外顯閱讀系統,並且以桃園市的一所大學的學生參加了這項研究,並且收集到了22例具有可信度的結果。我們將這些結果與事先製作好的參考答案做比較,並計算出Precision、Recall、Accuracy以及F-measure,並且挑出幾個切割方式比較不一致的句子觀察切割方式與英文小考分數的差異,並觀察學生所切割出來的語塊與詞性的關聯。分析後的結果顯示:(1)因為資料量不足,無法看出Precision、Recall、Accuracy以及F-measure 與英文能力之間的相關性。(2)在剖析方式差異較大的句子中,正確剖析句子的組別平均分數較高。(3)學生所切割出來的語塊開頭的單字詞性以動詞與介係詞居多,結尾的單字詞性以動詞或是名詞居多。根據以上結果,建議未來進行實驗時可以朝以下方向改進:(1)增加實驗人數。(2)縮短剖析的文章長度。(3) 更為詳細的前測與後測。 關鍵字:語塊、PC平面、外顯閱讀
According to psychological linguistics, reading is a cognitive process, which includes chunking words of a sentence into meaningful groups of words, called chunks. In this paper, the readers use a so-called “explicit reading” system to expose the implicit chunking process in their minds. A reader can determine which segments of the sentence possess complete meanings. Such segmentation can be expressed by clicking between the words to break or un-break a line of words into several chunks. A group of hierarchical chunks will be built up by a series of leading indentations of these chunks. Two classes of college students in Taoyuan city participated in this study and twenty-two feasible cases are collected. At the end of this process, the chunking result is scored by the calculation of precision, recall, accuracy, and f-measure analysis with respect to a referenced answer. After investigation, Several findings can be concluded: (1) The data is not enough to found the relation between the calculation score of precision, recall, accuracy, and f-measure analysis and the quiz score; (2)Through the chunking results analysis of those sentences which the chunking result are diverse among students , students with more correct way of chunking also can get higher quiz score; (3) The part of speech and the chunking can be also correlated.Several suggestions are proposed for future researches: (1) The number of experiments should be increased; (2) The article for segmentation should be shorter; (3) The pretest and posttest should be more detailed. Keyword: chunk, PC plane, explicit reading