| Summary: | 碩士 === 國立臺北科技大學 === 電機工程系 === 107 === With the popularity of technology uses, more and more work has been replaced by computers, and automatic summary systems have emerged and applied in many different situations. Taking the teaching scene as an example, teachers rely on past teaching experience to provide the key points of students' textbooks. If the computer program can automatically capture the focus, you can save a lot of time. Allowing teachers to tell more about the course content in a time-limited classroom to improve the quality of teaching.
Although automatic summaries have been developed for many years, there have been few automatic summaries in the past for textbooks. This thesis first introduces the textbook into the automatic abstract, and according to the more refined characteristics of the textbook, the algorithm proposed by Luhn is improved. First, we first retrieve the historical keywords defined by the Political University and the Ministry of Education, and then include these keywords in the system. Then, according to the number of keywords in the sentence, calculate the significant factors of each sentence through the formula. Value, and finally the sentence according to the level of the significant factor, to generate automatic summary with compression ratio of 30%, 20% and 10%, and carry out related evaluation work.
In this thesis, we have found “student” and “related experts” to evaluate the system. According to the assessment, the accuracy of the “unit” is too large, and the accuracy and recall rate are too large and extreme in some units. In this case, this shows that the assessment will be inaccurate due to the number of subjects and the knowledge of the subjects. In the ROUGE-1 evaluation, whether it is a student or expert evaluation, the results of each compression rate are fairly stable. It can be seen that this automatic summary can cover some important contents more objectively.
|
|---|
