한국해양대학교

Detailed Information

Metadata Downloads

한국어 서답형 자동채점에 적용하기 위한 형태소 분석 및 품사 부착

Title
한국어 서답형 자동채점에 적용하기 위한 형태소 분석 및 품사 부착
Alternative Title
Morphological Analysis and Part-of-Speech Tagging for Applying Korean Automated Scoring of Short-Answer Questions
Author(s)
천민아
Publication Year
2016
Publisher
한국해양대학교 대학원
URI
http://kmou.dcollection.net/jsp/common/DcLoOrgPer.jsp?sItemId=000002234039
http://repository.kmou.ac.kr/handle/2014.oak/10538
Abstract
Through Korean short-answer questions, we can reflect the depth of students’ understanding and higher-order thinking skills, but may take long time to grade and may be an issue on consistency of grading. To alleviate the suffering, automated scoring systems are widely used in Europe and America, but are in the initial research stage in Korean. Many language modules like morphological analysis are used to improve Korean automated scoring.

The previous morphological analyzer used for Koran automated scoring under development suffers from some unusal words like “우오오오오오오오오오오오오오오오오오오오오.” In this thesis, we propose a new method for Korean morphological analysis to solve this problem. The proposed method is combined with syllable-based word segmentation and versatile searching for morphological variants. The syllable-based word segmentation is based on a machine learning model like conditional random field (CRF) and based on the BIO coding scheme. The versatile searching for morphological variants comprises four steps: The first and second steps are to look up segmented words in the pre-analyzed dictionary and morphological dictionary, respectively. For unknown words, the third step is to search for the segmented word in the variant dictionary and to concatenate the variant words with the previous words and the next words. The final step is to look up the combined words in the morphological dictionary. At each step, words in the dictionary are added into nodes on the lattice structure , which is used for POS tagging and a weighted graph. The POS tagging is the best (shortest) path, i.e., the most proper sequence for a given sentence, from the beginning node to the last node on the weighted graph.

The proposed morphological analyzer and POS tagger has demonstrated the recall and of 98.86% and the precision of 95.03% for the SEJONG corpus, and also can analyze all answers of subjects taken the 2014 National Level Student Assessment. Thus it can be said that the proposed systems are more effective than the morphological analyzer and POS tagger used for the automated scoring system of Koran short-answer questions.
Appears in Collections:
컴퓨터공학과 > Thesis
Files in This Item:
000002234039.pdf Download

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse