The tool for term-segmentation test based on latent semantic analysis and its application in reading comprehension

• Main Authors: Wu, Chih-Ning, 吳芷寧
• Other Authors: Kuo, Bor-Chen
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/00377571493073258897

碩士 === 國立臺中教育大學 === 教育測驗統計研究所 === 101 === The purpose of this study is to use Latent Semantic Analysis (LSA) to define the familiarity of terms and to assess the effectiveness of the familiarity of terms according to the scores given by experts. Moreover, this study also exploits the advantage of LSA that can measure the correlation between each term to develop an auxiliary tool of the LSA term-segmentation test paper, and then evaluates the effectiveness of the LSA term-segmentation test generated by the auxiliary tool of the LSA term-segmentation test paper. Furthermore, this study also explores the effectiveness of the following four methods in predicting the reading comprehension and the amount of literacy: the traditional expert term-segmentation test, the LSA term-segmentation test, the traditional binary scoring system, and the LSA automated scoring system. In this study, the 5th grade students in the elementary schools in Taichung City were taken as the research subjects, and a total of 49 valid samples were collected. The research results are listed as follows: 1. There was a significant correlation between the familiarity of terms defined by LSA in this study and the scores given by experts, suggesting that the effectiveness of the assessment is good. 2. There was a significantly high correlation between the LSA term-segmentation test generated by the auxiliary tool of the LSA term-segmentation test paper that developed in this study and the traditional expert term-segmentation test, suggesting that the effectiveness of the assessment is good and the auxiliary tool of the LSA term-segmentation test paper can effectively save the time from developing a test paper manually. 3. The total explanatory power and individual explanatory power of the traditional expert term-segmentation test and the LSA term-segmentation test for the reading comprehension and the amount of literacy were compared in this study, and the analysis results show that the LSA term-segmentation test can effectively predict the reading comprehension and the amount of literacy. 4. The traditional binary scoring system and the LSA automated scoring system were compared in this study, and the analysis results show that the LSA automated scoring system has a higher predictive power for the reading comprehension.