在實作本系統時，我們將調整文章評分過濾器中所用到的門檻值，來達到適性化。我們也利用實驗數據來證明我們所訂定的門檻值是否能符合我們的系統，我們將六等級及資料庫中的資料，調整不同的門檻值來檢視六等級之分佈是否在資料庫中，在將之轉換成高斯分佈，看看兩者高斯分佈是否貼近，並以數據量化高斯分佈圖形，以數據檢視兩分佈是否接近，月接近表示兩者難易分佈是一樣的，而2000這個門檻值也的確是可以符合我們的系統的，這也是本論文之貢獻之一。 In this paper, we propose a Document Recommendation System on WWW for English as Second Language(ESL) learners. Actually we can say the system is a personal recommendation system for ESL learners.
First, we provide the similar pages for those learners, and the similar pages that we provide can be regarded as the same theme which the pages discussed. We also record the degree of learners and use a Language Difficulty Filter(LDF) to filter out the pages which is not accord with the learner’s degree as our main component in the system. The main idea of our system is to raise the rate of repeated exposure of the word which user wants to know. So we provide this system, in addition to raise the rate of repeated exposure of words, we also choose the pages which accord with the learner’s degree.
To test the actually system, we adjust the threshold for our system. With this system, we will build the Gaussian Distribution for both scores of six degree and data in our data base and then we will examine the Chi-Square test statistics from the distribution of them for the different threshold of LDF subsystem. After examining and analyzing the results, we concluded through expand by sense , the threshold (2000) of the LDF subsystem as a whole has a dramatic improvement of personally recommend. Beside the data in our data base, we can also use the keywords with a closer definition with the image we desire.
