| Summary: | 碩士 === 南臺科技大學 === 資訊工程系 === 107 === With the advancement of technology, the government has attached great importance to information education and programming in recent years. The Ministry of Education also expects to include such courses in the curriculum. However, with traditional paper textbooks and teaching methods, it is impossible to keep up with the ever-changing pace of technology. There is a lot of new knowledge, information needs to be obtained through the Internet, but the speed of information generated by the Internet is extremely amazing, and irrelevant information. There are also a lot of data with different levels of difficulty, which will cause learners to have a lot of inconvenience and waste of time when they look at themselves. Therefore, this study establishes an article retrieval system in order to provide suitable articles for learners. The system will first collect information type articles by web crawler, and then use TF-IDF (term frequency-inverse document frequency), association rules, etc. to explore the keyword correlation, similarity and other characteristics in the technical analysis information type article. And with these characteristics to establish the relevance between the articles, and finally according to the degree of similarity between the articles, hierarchical classification, and can be recommended to learners according to the relevance of each article.
This study also uses the test information type of the exam questions as the data, using the method used in the article search method to identify the difficulty to test the accuracy of the methods in this study, because each question is equivalent to a streamlined article, and The test questions can be set according to the answering situation, so the accuracy of the above method can be quickly determined.
The pictures existing in the articles and test questions are also an important basis for judgment, but the types of information type pictures are too broad, and the manual marks require a lot of manpower. For this study, the label of the pictures is set by the picture context... The method is also based on the function of capturing keywords to achieve automatic labeling of pictures. This method does not require a lot of cost for pre-processing and training on the classified pictures compared to the neural network.
|
|---|
