英語文試題檢測與答題驗證系統設計與實作

Abstract

在近年來全國不僅開始總動員學習英語，並且掀起了一股英語檢定風，造成統計至九十三年底，共有三十二萬人次參加全民英語檢定熱潮。在如此繁重的試務工作中，要如何讓檢定具有一定水準，讓試題的分級有一可靠的機制及參考標準，讓批改答題的過程更有效率，這些都是我們所關心的，而且深深的影響到檢定單位的公信力。 因此本論文嘗試去研究電腦化英語檢定測驗過程，如何讓試題列入適當的檢定級數，又如何能讓電腦來協助閱卷老師，分擔閱卷批改過程的負荷，進而使得評分閱卷老師的閱卷基礎較一致。進行試題檢測是因為要達到試題分級的目的，因此我們提出利用詞彙表、字頻及詞彙數分析的方法，在命題委員命題完畢後，能夠經過試題檢測客觀的分析，讓分級的過程中，有一個分級評量標準，使得試題分級更具有公信力。在答題驗證部份，我們實作出拼字檢查、文法檢查及語義檢查三大部份，輔助閱卷老師在批改過程中的各項檢查，以減少人工閱卷之時間及錯誤。

The English language skill training (teaching), learning, and assessment is a big industry around the world, especially in the territory of Asia countries. The traditional pen-paper based test delivery approach for English language skill assessment is still domination. However, the computer based test delivery approach has gradually received its attentions due to its advantages on using multimedia such as video, animation, voice, and color image to simulate the question presentation in the real world. Also, using computer and network to conduct a test delivery is less labor intensive and is relative efficiency for the test delivery management activities. In any test delivery approaches, test questions have to be designed and validated and answers from test takers must be graded. From test management point of views, each pen-paper based test sheet requires a qualified score marker to grade it. The grading process usually is very time-consuming. Considering that there are million copies of test sheets needed to be graded, a lot of qualified score markers are demanded. And, usually, the grading criteria of each score marker may not be able to retain a consistent way after a long grading hour. In this thesis study, we attempt to design and implement 1) an English test question validation system and 2) a score marking system for English writing test under a computer based test delivery approach. Especially, the proposed English test question validation system is designed to help test question creators to verify if a created test question is met with the level of difficulty by analyzing several test item attributes associated with the question. Based on the report generated from the test question, test question creators can adjust the level of difficulty before the test question is submitted to test question bank. The proposed scoring marking system for English writing test is designed to help score marker to give a more consistent and accurate grading result by analyzing the vocabulary, grammar, and possibly semantic errors found in the answer sheet. Each score marker can have a report analyzed by the proposed system for reference before a formal score is given. This will avoid problems mentioned above. We integrate several current tools (such as Aspell, Apple Pie parser, Worldnet, and so on) into our score marking system and test question validation system. The proposed system has been used to the NETPAW English test to demonstrate its applicability and feasibility.