標題: 非監督式中文寫作自動評閱系統
An Unsupervised Chinese Automated Essay Scoring System
作者: 陳彥宇
李嘉晃
網路工程研究所
關鍵字: 自然語言處理;寫作自動評閱;中文寫作;natural language processing;automated essay scoring;Chinese essay
公開日期: 2006
摘要: 寫作自動評閱技術在語文教育領域中是十分重要的輔助工具,然而目前的自動評閱系統均需要一定數量的同主題文章做為訓練資料,因而在使用上有其限制性。本論文提出一個基於文章相似度的中文寫作自動評分系統,此系統不需要同主題文章的人工評分資料,僅需一定數量的測試資料,便可藉文章間的相似度及其表面特徵資訊自動進行評分。另外,本論文亦提出一個演算法可以偵測不合題旨的文章,以避免評分的錯誤。實驗結果允許一級分誤差的正確率可達95%,完全命中的精確率可達52%;可以做為人工評分的參考依據。
Automated essay scoring is an important research in education domain, but the cost of training is a problem. In this paper, we describe a Chinese automated essay scoring system based on similarity between test essays. This system could grade essays without graded training data. We also describe an algorithm detecting off-topic essays. The adjacent rate in the experiment is about 95%, and the exact rate is about 52%.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009456502
http://hdl.handle.net/11536/82168
Appears in Collections:Thesis


Files in This Item:

  1. 650201.pdf

If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.