intelligent database systems lab n.y.u.s.t. i. m. development of a reading material recommendation...

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

Development of a reading material recommendation system based on a knowledge engineering approach

Presenter : Yu-hui Huang

Authors :Ching-Kun Hsu, Gwo-Jen Hwang , Chih-Kai Chang

CE 2010

國立雲林科技大學National Yunlin University of Science and Technology

1


N.Y.U.S.T.

I. M.Outline

Motivation

Objective

Methodology

Experiments

Conclusion

Comments

2


N.Y.U.S.T.

I. M.Motivation

It is importance of assigning proper articles to individual students for training their reading ability.

In traditional English classes, a teacher needs to guide dozens of students to learn;

therefore, it is quite often that identical instructional materials, especially reading articles, are prepared for every student .

For some students, the articles could be too easy to read, while for others, the articles might be too difficult.

Such an article assignment strategy is likely to cause the students to lose interest in learning English

3


N.Y.U.S.T.

I. M.Objective

To improve the reading performance of students, it is important to provide personalized reading recommendations to individual students by taking their profile or learning performance into consideration.

A knowledge engineering method is proposed to assist teachers to cooperatively define English article recommendation rules for individual students.

4


N.Y.U.S.T.

I. M.Methodology-Establishing repertory grids

A single repertory grid is represented as a matrix whose columns have element labels and whose rows have construct labels. Elements

Elements could be decisions to be made, objects to be classified, or concepts to be learned.

Constructs are the features for describing the similarities or differences among the elements.

5


N.Y.U.S.T.

I. M.Methodology

Each construct consists of a trait and the opposite of the trait. A 5-scale rating mechanism is usually used to represent the relationships between the elements and the constructs. ‘‘5” represents that the element is very likely to have the trait;

‘‘4” represents that the element may have the trait;

‘‘3” represents ‘‘unknown” or ‘‘no relevance”;

‘‘2” represents that the element may have the opposite characteristic of the trait;

‘‘1” represents that the element is very likely to have the opposite characteristic of the trait in the study.

6


N.Y.U.S.T.

I. M.Methodology

In the study, the expert knowledge was obtained by interviewing two experienced English teachers in a senior high school.

Furthermore, the features and preferences for characterizing the learners were acquired from both the students and the experts.

That is, two kinds of repertory grids were established. One is the repertory grid for categorizing the selected reading articles;

the other kind is for characterizing the learners based on their preferences for English readings.

7


N.Y.U.S.T.

I. M.Methodology

For the first kind of repertory grid developed by interviewing the teachers, the data concerning the articles with different difficulty levels were collected and presented in the corresponding grids.

The English teaching experts determined the difficulty levels of 100 articles based on the vocabulary and sentence difficulty degrees as defined by the General English Proficiency Test (GEPT).

8


N.Y.U.S.T.

I. M.Methodology

For the second kind of repertory grids, the preferences of individual students for English readings are collected and recorded.

9


N.Y.U.S.T.

I. M.Methodology-Fitness analysis

A fitness analysis formula is used to compare the preferences of individual students with the traits of each article. N is the number of constructs (or traits),

MaxScore is the maximum rating in the repertory grid,

Ai represents the ith article, Sj represents the jth student

|gi,k gj,k| represents the distance between the ith article and the jth student based on the kth trait in the repertory grid.

10


N.Y.U.S.T.

I. M.Methodology-Fitness analysis

11


N.Y.U.S.T.

I. M.Methodology-Article recommendation strategy

Based on the fitness analysis results, an English article recommendation strategy is proposed for developing the expert system: Step 1: Identify the English reading ability of the student.

Step 2: Determine the candidate list of articles based on the English reading ability of the student.

Step 3: Test whether the condition |gic gjc| < 3 is true.

Step 4: If |gic - gjc| < 3, calculate the fitness degree between the student preference and the topic of the article.

Step 5: If |gic - gjc|>=3, test whether the condition (gic - gjc) < 3 is true.

Step 5.1: If (gic - gjc) < 3, calculate the fitness between the student preference and the topic of the article.

Step 5.2: If (gic - gjc)>=3, no recommendation will be given and fitness = 0.

12


N.Y.U.S.T.

I. M.Experiments

13


N.Y.U.S.T.

I. M.Experiments

14


N.Y.U.S.T.

I. M.Experiments

15


N.Y.U.S.T.

I. M.Experiments

16


N.Y.U.S.T.

I. M.Conclusion

1717

The major contribution of this study is to propose a way for developing an expert-like English reading recommendation by taking both preferences and knowledge levels of individual students as well as categories and traits of articles into consideration.

In the future, it is suggested that the number of articles should not only be continually increased to promote the chances of successfully matching the users’ needs, but that the number of article categories should also be increased so that more students can get recommended articles.


N.Y.U.S.T.

I. M.Comments

Advantage …

Drawback …

Application …

18

intelligent database systems lab n.y.u.s.t. i. m. development of a reading material recommendation...

Documents