臺灣博碩士論文加值系統

隨著科學技術的發展和進步，學習模式也在不斷發展。在問題驅動的學習中，學生通過回答問題來澄清和驗證他們學到的知識。如此眾多的問題需要良好的管理。良好的管理可以避免由於模糊表達而將具有相同知識集的學習材料定義到不同部分的情況。在篇論文中，我們提出了一個專注於國中小數學題目的階層式分類系統。我們測試了不同文件表示和分類器的幾種組合，其中包括扁平多標籤分類和知識點的階層式多標籤分類。針對包含文本和圖像的當前問題，我們還提出了文字與影像的預處理方法和組合文字影像特徵的CNN分類模型。實驗表明，我們的階層式分類可以勝過扁平多標籤分類方法。

With the development and progress of science and technology, the learning patterns also evolve. In Question-Driven learning, students clarify and validate what they learn by answering questions. Such a large number of questions needs good management. A well-performed management can avoid the situation that learning materials with the same knowledge set are defined into different sections due to ambiguous expressions. In this work, a hierarchical classification system that focuses on K-12 learning materials is proposed. We test several combination of document representation with flatten label and hierarchical label of the knowledge points. In response to a current question that contains text and image, we also introduce a multi-modal preprocessing approach and a combined feature CNN model. The experiment shows that our hierarchical classification can outperform flatten multi-label classification methods.

摘要 I
Abstract II
致謝 III
Table of Contents IV
List of Figures VII
List of Tables VIII
1. Introduction 1
2. Related Works 5
2.1 Question Classification 5
2.2 Multi-model Document 7
2.3 Hierarchical Multi-Label Classification 9
2.4 Summary 10
3. A hierarchical Multi-label Classification System of K-12 Cross-Knowledge Points Math Question Documents 12
3.1 Preprocessing 13
3.1.1 Image preprocessing 13
3.1.2 Mathematical Expressions Preprocessing 15
3.1.3 Text Noise Elimination Preprocessing 16
3.1.4 Tokenization 16
3.2 Document Representation 17
3.2.1 TFIDF 17
3.2.2 Word2Vec and Text Map 18
3.2.3 Convolutional Neural Network for image feature 20
3.2.4 BERT 21
3.3 Classification 22
3.3.1 Binary Relevance SVM, Binary Relevance XGBoost 22
3.3.2 LSTM classifier 22
3.3.3 Hierarchical local classification 24
Training 24
Predicting 25
4. Experiment 29
4.1 Data Set 30
4.2 Data Setting 33
4.2.1 Hierarchy tree pruning 33
4.2.2 Labeling policy 34
4.3 Compared Methods 35
4.4 Evaluation Metric 36
4.5 Experiment Result 37
4.5.1 Flatten multi-label classification 37
4.5.2 Hierarchical local classification 39
5. Conclusion & Future Work 42
List of Reference 43

[1] Stephen J. H. Yang, Jeff C. H. Huang and Anna Y. Q. Huang, “MOOCs for K-12 and Higher Education in Taiwan,” 2016 5th IIAI International Congress on Advanced Applied Informatics, IIAI-AAI, 2016
[2] Nagi, Jawad, et al, “Convolutional neural support vector machines: hybrid visual pattern classifiers for multi-robot systems,” Machine Learning and Applications (ICMLA), 2012 11th International Conference on, vol. 1, 2012.
[3] Niu, Xiao-Xiao and Suen, Ching Y, “A novel hybrid CNNSVM classifier for recognizing handwritten digits,” Pattern Recognition, vol. 45, no. 4, 2012, pp. 1318-1325.
[4] Peng Zhou, Zhenyu Qi, et al, “Text Classification Improved by Integrating Bidirectional LSTM with Two-dimensional Max Pooling”, arXiv:1611.06639, 2016.
[5] Dalal, Mita K., and Mukesh A. Zaveri., “Automatic text classification: a technical review.,” International Journal of Computer Applications 28.2, 2011, pp. 37-40.
[6] Tokinori Suzuki and Atsushi Fujii, “Mathematical Document Categorization with Structure of Mathematical Expressions”, 2017 ACM/IEEE Joint Conference on Digital Libraries (JCDL), 2017
[7] La Cascia, Marco, Saratendu Sethi, and Stan Sclaroff, “Combining textual and visual cues for content-based image retrieval on the world wide web,” Content-Based Access of Image and Video Libraries, 1998. Proceedings. IEEE , 1998.
[8] Denoyer, Ludovic and Vittaut, Jean-Noel and Gallinari, Patrick and Brunessaux, Sylvie and Brunessaux, Stephan, “Structured multimedia document classification,” Proceedings of the 2003 ACM symposium on Document engineering, 2003, pp. 153-160.
[9] Seo, Min Joon, et al., “Diagram Understanding in Geometry Questions,” AAAI., 2014.
[10] Huang, FJ and LeCun, Y, “Large-scale learning with svm and convolutional nets for generic object recognition,” 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2006.
[11] Yannis Papanikolaou, Ioannis Katakis, Grigorios Tsoumakas, “Hierarchical Partitioning of the Output Space in Multi-label Data”, arXiv:1612.06083, 2016
[12] Jacob Devlin, Ming-Wei Chang, Kenton Lee, Kristina Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding”, arXiv:1810.04805, 2018
[13] Carlos N. Silla Jr., Alex A. Freitas, “A survey of hierarchical classification across different application domains”, Data Mining and Knowledge Discovery , 22(1-2), 31 - 72, 2011
[14] Shao-Yi Ho, “A Self-Relevant CNN-SVM Model for Problem Classification in K-12 Question-Driven Learning”, July 2016
[15] Jyun-Kai Chen , “CNN-SVM Hybrid Classifier: Multi-label Classification in K-12 Cross-chapter Problem”, July 2017