應用文本分類提供以概念圖為基礎的學習資源推薦;Applying Neural Network to provide learning resources recommendation based on concept map

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/77529

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/77529

Title:	應用文本分類提供以概念圖為基礎的學習資源推薦;Applying Neural Network to provide learning resources recommendation based on concept map
Authors:	顏承印;Yen, Cheng-Yin
Contributors:	資訊工程學系
Keywords:	文本分類;深度學習;卷積神經網路;多層感知器;長短期記;支持向量機;Text Classification;Deep Learning;Convolution Neural Network;Multi-Layer Perceptron;Long-Short Term Memory;Support Vector Machine
Date:	2018-07-13
Issue Date:	2018-08-31 14:47:14 (UTC+8)
Publisher:	國立中央大學
Abstract:	本研究以ilearning線上學習平台為基礎，透過蒐集該平台上之課綱、電子教材、簡報教材當作資料來源，進行文本分類，並依據概念圖來實現適性化學習教材推薦，提供授課教師進行介入干預。教材之推薦依據將透過卷積神經網路(Convolution Neural Network，CNN)、多層感知器(Multi-Layer Perceptron，MLP)、長短期記憶(Long-Short Term Memory，LSTM)與支持向量機(Support Vector Machine，SVM)等幾種神經網路與傳統機器學習的方式建立章節分類模型，同時結合一個Enrichment Method解決教材文本數量缺少、內文不足的情況，希望透過此短文本分類模型，結合Blended Course課程，分析出授課教師在設計課綱及課程教材上的規劃思路，藉此建立智慧分類模型。文本分類準確度將以accuracy、precision、recall及f1-score做為評估指標，實驗結果顯示，透過本研究的Enrichment Method處理過後，各個指標在大部分的分類模型上皆有所提升，其中又以CNN的進步比例最多，達到30%，而整體成效以MLP表現最好。實驗對象經由本推薦流程的介入之後，透過T檢定測試，發現在進步成績上，有推薦的實驗組進步成績相對於無推薦的控制組進步成績是有顯著差異的。;Based on the ilearning online learning platform, this study collects syllabus, online materials on the platform as source of data, to classify context of webpage, and implements appropriate learning recommendations based on concept maps, providing interventions. The recommendation is rely on the results of classification model build by several deep learning neural network classification models, which comprise Convolutional Neural Network, Multi-Layer Perceptron, Long-Short Term Memory and Support Vector Machine. We used a data enrichment method to solve the issue of lack of samples and insufficient text. We want to combine these short text classification models and the blended course (operating system) to analyze the idea of instructor in designing course syllabus so as to establish a wisdom classification model. In this study, we used accuracy、precision、recall and f1-score to evaluate the outcome of classification model. The experimental results show that after the Enrichment Method, most of the classification models have improved, where CNN has the highest proportion of progress, reaching 30%, and MLP has the best performance in overall. With our recommendation, students’ progress has significant difference between control group and experimental group through T test.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	145	View/Open

社群 sharing

Loading...