Development of a statistical information management system for large scale test data

在这项研究中，我们开发了一个组织试题、考生答题数据和考生个人资料的信息管理方案。基于该方案，我们开发了一个测试信息管理实验系统。1)研究了各种基于XML2的码本描述标准化方案。2)在具有高度符号操作能力的逻辑编程语言Prolog上开发了一个用于处理基于XML的码本的函数库。3)在关系数据库管理系统(RDBMS)上实现了一组测试题和相关的测试成绩，并对其在大规模统计数据上的处理性能进行了评估。4)设计了一种表示测试项与由测试响应得到的统计数据之间关系的方法。这些统计量由基本统计量、各种条件统计量和基于IRT(项目反应理论)的统计量组成。5)对一个基于n元文法的文本检索系统进行了测试。研究了考试项目之间关系的管理方法和考试成绩的统计信息。6)利用开发的系统对部分国家中心考试的考试成绩进行了分析，指出了考试成绩的一些重要特点。

In this research, we developed an information management scheme for organizing test items ; examinees' answer data, and examinees' profiles. Based on this scheme, we developed an experimental system for test information management.1) We surveyed various proposals for standardization on codebook description schemes with XML2) We developed a function library for handling XML based codebooks on Prolog, a logic based programming language with high symbolic manipulation abilities.3) We implemented a set of test items and related test scores on an RDBMS (relational database management system), and evaluated its processing performance on large scale statistical data.4) We designed a method for representing the relationship between test items and statistical data that was derived from test responses. These statistics are composed of basic statistics, various conditional statistics, and IRT (item response theory) based statistics. We implemented the representation scheme on a RDBMS.5) We evaluated an n-gram based text retrieval system for test item management. Methods for managing relationship between test items and statistical information on test scores are examined.6) We analyzed test scores of a part of the National Center Test and pointed out some important features with the aid of the developed system.