도움말

A Stage Transition Model for Korean Part-of-Speech and Homograph Tagging

Vol.39 No.11, 2012.11, 889-901 (13 pages)
Copy
Quick View Quick View
Purchase $5.29
Export
Usage : 258
Citations : 1
분야내 활용도 : 5%
More detail >

· Usage : Full-text article downdloads count since 2010.

· Citations : Cited in the DBpia's articles

· Impact Score : Calculates the article impact score on a basis of the usage in the last 24 months.

Abstract
한국어 의미처리시스템은 어절의 형태소를 분석하고, 품사와 동형이의어를 분별해야 한다. 이를 위해 최근의 연구들은 말뭉치를 기계학습하여 품사 및 동형이의어 태깅하는 방법들을 제안하고 있으며, 기계학습에서 자료부족 문제를 해결하는 것에 집중하고 있다. 본 논문에서는 말뭉치를 기계학습하여 품사와 동형이의어를 동시에 태깅하는 방법을 제안한다. 이 방법은 자료부족 문제를 해결하기 위해 단계별 전이모델을 사용하며, 각 전이모델마다 다른 가중치를 적용한다. 또한 부분적으로 전이빈도가 0인 경우를 위해 최소 전이점수를 계산하는 예외 처리 루틴을 포함한다. 실험을 위해 1천만 어절의 세종 말뭉치를 사용하였으며, 이 중 90%를 학습하고 나머지 10%를 테스트하였다. 실험 결과 품사와 동형이의어 둘 다에 대해 96.44%의 태깅 정확률을 보였으며, 품사만을 고려할 경우 98.0%의 태깅 정확률을 보여 본 논문에서 제안하는 단계별 전이모델이 한국어와 같은 교착어의 품사 및 동형이의어 태깅에 적합한 방법임을 알 수 있다.

A Korean semantic processing system must analyze morpheme and disambiguate the POS and homograph for words. The recent researches proposed machine learning model using corpus, and focused on resolving the data sparseness problem. This paper proposes a method of simultaneous tagging for the POS and homograph using machine learning of corpus. The proposed method uses stage transition model for resolving the data sparseness problem, and applies different weight to each transition model. The tagging system includes exception routine that calculates minimum transition score in the case of no transition between words. For the experiment we used the 10 million words Sejong corpus and learned 90% of the Sejong corpus and tested 10%. The experiments shows that the accuracy is 96.44% for the POS and homograph tagging and 98.0% for the only POS tagging. From the experiment, the proposed stage transition model is adequate for tagging POS and homograph of agglutinative language like Korean.

TOC
요약
Abstract
1. 서론
2. 관련 연구
3. 단계별 전이모델
4. 구현 태깅시스템(UTagger)
5. 실험과 결과 분석
6. 결론 및 향후 연구 방안
참고문헌
Keyword
References (23)

Please found references of this article.

  1. Dong Myung Kim , 2009 , Simultaneous Korean POS and Homonym Tagging System using HMM , 석사 , Ulsan University

  2. Myeong-jin Hwang , 2007 , A Dynamic Link Model for Korean POS-Tagging , Proc. of Conference of Hangul and Korean Information Processing : 282 ~ 289

  3. Jung-ho Shin , 1994 , An HMM Part-of-Speech Tagger for Korean Based on Wordphras , Proc. of Conference of Hangul and Korean Information Processing : 389 ~ 394

  4. Young-Hoon Kim , 2002 , An Efficient Korean Part - of - Speech Tagging , Journal of the Korea Contents Association 2 (2) : 98 ~ 102

  5. Jae-Hoon Kim , 1995 , An Efficient Korean Part - of - Speech Tagging using a Hidden Markov Model , Journal of KIISE 22 (1) : 136 ~ 146

  6. 박희근 , 2007 , 어절별 중의성 해소 규칙을 이용한 혼합형 한국어 품사 태깅 시스템 , 정보과학회논문지 : 컴퓨팅의 실제 및 레터 13 (6) : 427 ~ 431

  7. Yuhwan Kang , 1999 , Korean Part- of-Speech Tagging using Constrained-Rule and Main POS Information among Words , Proc. of the 11th Conference on Hangul and Korean Language Information Processing : 433 ~ 437

  8. Scott M. Thede , 1999 , A Second-Order Hidden Markov Model for Part-of-Speech Tagging , Proc. of the 37th of ACL : 175 ~ 182

  9. Julian Kupiec , 1992 , Robust part-of-speech tagging using a hidden Markov model , Computer Speech and Language 6 (3) : 225 ~ 242

  10. Eugene Charniak , 1993 , Equations for part-of- speech tagging , Proc. of the Eleventh National Conference on Artifical Intellige : 784 ~ 789

  • 처음
  •  
  • 이전
  •  
  • 1
  •  
  • 2
  •  
  • 3
  •  
  • 다음
  •  
  • 마지막
Cited articles (1)

Please apply for Alerts and check the information by e-mail!

Other articles of first author (10)

Please check the detailed of Joon-Choul Shin Identified author!

Within this Journal (10)

Please check the detailed of Vol.39 No.11!

Recommended Articles (10)

We provide services, 'DBpia Recommended Articles' and 'Customers who used this article also used', that used text mining, usage and citations data.

DBpia Recommended Articles

More recommended articles!

Customers who used this article also used

Metrics

Usage Status

· Usage

· Top 3 institutions list on usage

More detail >
No Top institutions Usage
1 충북대학교 20
2 서울대학교 14
3 연세대학교 14

Impact Score

· Impact Score

· The article impact score on the subject

More detail >

: %

2016-09
2016-10
2016-11
2016-12
0
20
40
60
80
100
  • 0%
  • 20%
  • 40%
  • 60%
  • 80%
  • 100%

Citations

Detailed Info
Copyright Policy

The copyright of all work are belongs to the original author. The contents of each work shall not be responsible or guarantee. Crawl the metadata of articles do not allowed without agreement.

top