English abstract
Education is a critical factor in the long-term economic development of any society.
There are many learning materials that instructors and students use to learn and find
knowledge for learning purposes in the academic sector. These learning materials
could be in different formats (i.e., PDF files, reports, Case studies, Manuals, etc.).
Students and instructors may have difficulty finding information in these materials
related to a specific course due to the tremendous amount of information
presented. Therefore, this research proposes A Framework for a Guided Extractive
Summarization of Learning Content (GESLC) to summarize the course's learning
content to achieve the course objectives. The main contribution is proposing and
developing a novel framework that combines several deep learning algorithms to
provide efficient summarization techniques to summarize the learning content
according to the course outline. This Framework consists of four layers. The first
layer is the input layer, which will take the course outline and the material content
for the selected course. The course outline is analyzed to match the optimal format
that describes course objectives and other related information. Second, the primary
key is extracted from the outline after identifying the keywords/key phrases.
On the other hand, the course materials are preprocessed to be used as input for the
summarization algorithm. The summarization layers consist of query-based and
deep-learning algorithm layers. First, query-based summarization is applied to
produce the candidate sentences according to the keywords extracted from the course
outline that match the course's objectives. Finally, the deep learning algorithm – The
restricted Boltzmann Machine algorithm-summarizes a candidate's content according
to well-defined criteria to produce the final summary.
Several methods are utilized in this study to evaluate the proposed Framework. First,
statistical evaluation is applied using Recall-Oriented Understudy for Gisting
Evaluation (ROUGE), the most popular way to evaluate automatic text
summarization. This evaluation calculates three metrics: Precision, Recall, and F Score. The summarized content is considered against the human-based
Summarization. The second method evaluation method was human evaluation. The
summarized content was given to some experts and evaluated according to some
criteria; the third method of evaluation used case studies from an actual educational
environment. Two courses are selected from the SQU environment. Both courses
contain an adequate number of students registered in one semester. The evaluation
process shows better results in guiding instructors and students to summarize
learning content according to the course objectives to finally have a perfect summary
matching the learning process's objectives and enhancing student achievement.
Furthermore, the evaluated methods improve the precision of the state of art for
extractive summarization.
