近年來網際網路的蓬勃發展,帶來了學習方式的革新,有別於傳統課堂的數位學習模式日趨盛行,現今越來越多人使用數位學習的方式進行學習。評估學習效果時:傳統課堂主要以考試作為學習成效評估;而數位學習亦常使用線上題庫測驗來評估學習成效。但以目前的線上題庫系統,其自動出題無法直接判斷比對所出題試之間的相似度,因此本研究旨在於應用潛在語意分析方法(Latent Semantic Analysis, LSA)評估題目與題目之間相似程度,希冀能此分析方法予以未來線上題庫測驗出題之事前評估,達到有效避免出現試題答案出現相似或觀念雷同試題之目的。 本研究測試結果發現: (一) 實驗設計中對於斷詞後去除冗詞與未去除冗詞,在試題判斷相似度上,去除冗詞在判斷相似度上較佳。 (二) 物流關鍵字在一般的斷詞系統中尚未能有效進行物流關鍵字斷句,故本研究結果未得顯著性結果。 (三) 潛在語意分析對於計算相似度過程中,維度約化值的大小會影響試題判斷成效,維度約化值如何取決乃為實務應用的關鍵參數;相較於向量空間模型法(Vector Space Model, VSM)之結果言,潛在語意分析能達更精確判斷試題相似程度。 綜合上述研究結果可以證明,對於試題之相似程度分析,明顯得出潛在語意分析與向量空間模型兩者在比對試題相似度成效上,潛在語意分析更能精確判斷試題相似程度,是故潛在語意分析方法極具成為比對題庫試題相似程度之重要技術。
In recent years, the booming Internet has brought the new style of learning. Unlike the traditional learning, e-learning gets popular increasingly. Nowadays, more and more people are using e-learning approach. When evaluating the effect of learning, the traditional classroom uses paper and pencil tests to measure the learning effect. However, e-learning would use online exam tests to evaluate. Nevertheless, the current online exam system can not directly determine and compare the similarity between exam questions in the item pool. Therefore, this study is going to take advantage of Latent Semantic Analysis (LSA) to evaluate the degree of similarity between exam questions. To achieve the goal of avoiding the appearance of similar exam questions or concepts, I hope this analysis approach can be applied to the advance evaluation of setting online exam in the future. In this study, the results show that: (i) In the experimental design regarding removing verbiages or not, removing verbiages is a better way to judge the similarity. (ii) In general word segmentation system, logistics terminology is not able to conduct an efficient result yet. Therefore, the result is still without the significant outcomes. (iii) In the process of calculating the similarity with LSA, the magnitude of dimension reduction will affect the efficiency when judging the exam question. How the dimension reduction is depends on critical parameter for the practical application. Compared to the result of Vector Space Model (VSM), LSA can determine the similarity of exam questions more accurately. To sum up, with regards to the similarity analyze for exam questions by LSA and VSM, LSA could precisely determine the similarity. Consequently, LSA is a technique which is important for comparing the similarity of exam questions.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。