The Key Technologies Research On Automated Assessment Technology For Sujective Tests

Nowadays, with the high speed of information developing in the world, IT technology plays a very important role in peoples’ lives, and brings many benefits to us. The informatization and intelligentization of the exam assessment system is an important branch of the IT technology accessing to the education. By now, the exam assessment system can be able to do the automatic scoring on the objective tests. As for subjective tests, it is not useful. There are NLP, Artificial intelligence, pattern matching and other technologies involved in automatic scoring on the subjective items. It is very difficult to do it in Chinese.Automatic scoring on the subjective tests will calculate the similarity of the sentences, and the most important task is word segmentation, which is the basis of the automatic scoring. The accuracy of the word segmentation will affect the result directly. In this paper, we do the research of the Chinese word segmentation techniques in-depth, and then compare their advantages and disadvantages. In the research, we do the Chinese word segmentation by Lucene.The answers which the students give can not be totally same with the model answer, so we must calculate the similarity of model answer and the answers that the students give. The key part in the sentence similarity is the similarity between words, which will affect the efficiency and accuracy of the whole system. So the key point of the research is the word similarity in the basis of the word segmentation. The research is based on the word similarity in HowNet, which is a world knowledge dictionary about the relationship between words and concepts. Based on that, an improved algorithm is put forward, and takes the depth of the first sememe tree into consideration. In addition, we also do the experiment on the accuracy in different word order in the answers.The expertiment shows that the automatic scoring on the subjective tests can meet the requirement, also improve the accuracy of the automatic assessment for subjective tests.