对从HSK题库中计算机自动生成试卷稳定性的试验检验

An Experimental Test of the Test Form Generated Automatically by Computer from HSK Item Bank

由计算机从题库中自动生成的试卷能否保持难度的相对稳定？根据IRT进行的等值误差范围有多大？为了回答这些问题，本文以共同组等值作为标准，对基于IRT之上的共同题等值误差进行了试验检验。试验中，采取一定措施保证了考生的动机水平。结果显示，IRT等值的校正方向都是正确的。在4个分测验中有3个分测验的的等值校正效果较理想，1个分测验的等值校正效果不够理想。计算机自动生成的试卷与原有人工命制的试卷在得分方面比较一致，分数相关达到0.931，获得证书的情况也是比较一致的。

To testify the stability of the test form generated automatically by computer from HSK(Chinese Profi- ciency Test) item bank and to estimate the equating error based on item response theory (IRT), the author de-signed an experimental research.Certain measures had been taken to guarantee the motivation of the testee.The results of this research illustrate that IRT equating procedure decrease the measurement error in all 4 sub-tests.Errors of 3 sub - tests were largely corrected while error of 1 sub-test was only partly corrected. The score of the test form auboatically generated is consistent with the traditional artifidally edited test form. The correlationindex boween 2 test forms is 0.931. The qualified rates of the two tests are also similar.