文章摘要寫作評量系統 Summarization Scoring System

Date
2005
Authors
丁偉民
Wei-Ming Ding
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
本研究之主要目的為發展一國小摘要寫作評量系統。評量系統主要參考「潛在語意分析」(LSA)的方法,利用SVD (singular value decomposition)的技術來建立潛在語意空間,並建立不同文體及不同大小的語意空間,在不同的語意空間中,比對教師及學生的摘要的關鍵詞,作為摘要評量的依據。除此之外,我們也探討其他摘要評量的指標,希望能夠從中找出適合中文摘要評量的指標。 本研究以台北市西門國小五年級的三個班級為實驗對象。經過學生摘要寫作的實驗過程之後,評估經由系統所獲得的各評量指標與教師人工評量的成績,在不同文體及不同大小的語意空間中的相關性。研究結果有一下發現:(1)使用SVD轉換的技術來建立語意空間,並在語意空間中做比對,可達到不錯的評量效果。(2)本研究嘗試使用教師及學生摘要句子的比對方式來評量學生摘要寫作,發現是值得繼續研究的方向。(3)不同文體及不同大小的語意空間,對於評量指標會有不同的影響。
The main purpose of this research is to develop a summarization scoring system for the teacher of the elementary school. This system refers to a method called Latent Semantic Analysis(LSA), to use Singular Value Decomposition(SVD) to build the semantic space. We build several kinds of semantic spaces which the size and style of writing are different. To compare the keywords of the summarizations between the teacher and the students in the different semantic spaces for the scoring. In addition to scoring, we analyze the other indexes of the summarization scoring to find the more appropriate approaches to summarizing for Chinese. The participants are the students of Xi-men elementary school. After the processes of the experiments of the summarizing, we analyze the correlation between all kinds of indexes of scoring calculated by the system and the teacher of scoring in different kinds of semantic spaces. The Results of the research are :(1)We can get a good result when we compare the summarization between the teacher and the students in the semantic spaces built by the translation of the SVD.(2)We try to scoring the summarization written by the students by comparing to the sentences of the summarization between the teacher and the students, and think that this approach is a worthy aspect to research.(3)The difference of size and the style of writing will effect upon the analysis of the result for the indexes of the scoring.
Description
Keywords
LSA, SVD, 關鍵詞, 摘要, 摘要寫作評量, LSA, SVD, keyword, summarization, Summarization Scoring
Citation
Collections