• Title/Summary/Keyword: ALBERT

Search Result 229, Processing Time 0.027 seconds

Recent R&D Trends for Pretrained Language Model (딥러닝 사전학습 언어모델 기술 동향)

  • Lim, J.H.;Kim, H.K.;Kim, Y.K.
    • Electronics and Telecommunications Trends
    • /
    • v.35 no.3
    • /
    • pp.9-19
    • /
    • 2020
  • Recently, a technique for applying a deep learning language model pretrained from a large corpus to fine-tuning for each application task has been widely used as a language processing technology. The pretrained language model shows higher performance and satisfactory generalization performance than existing methods. This paper introduces the major research trends related to deep learning pretrained language models in the field of language processing. We describe in detail the motivations, models, learning methods, and results of the BERT language model that had significant influence on subsequent studies. Subsequently, we introduce the results of language model studies after BERT, focusing on SpanBERT, RoBERTa, ALBERT, BART, and ELECTRA. Finally, we introduce the KorBERT pretrained language model, which shows satisfactory performance in Korean language. In addition, we introduce techniques on how to apply the pretrained language model to Korean (agglutinative) language, which consists of a combination of content and functional morphemes, unlike English (refractive) language whose endings change depending on the application.

Hierarchical Learning for Semantic Role Labeling with Syntax Information (계층형 문장 구조 인코더를 이용한 한국어 의미역 결정)

  • Kim, Bong-Su;Kim, Jungwook;Whang, Taesun;Lee, Saebyeok
    • Annual Conference on Human and Language Technology
    • /
    • 2021.10a
    • /
    • pp.199-202
    • /
    • 2021
  • 의미역 결정은 입력된 문장 내 어절간의 의미 관계를 예측하기 위한 자연어처리 태스크이며, 핵심 서술어에 따라 상이한 의미역 집합들이 존재한다. 기존의 연구는 문장 내의 서술어의 개수만큼 입력 문장을 확장해 순차 태깅 문제로 접근한다. 본 연구에서는 확장된 입력 문장에 대해 구문 분석을 수행 후 추출된 문장 구조 정보를 의미역 결정 모델의 자질로 사용한다. 이를 위해 기존에 학습된 구문 분석 모델의 파라미터를 전이하여 논항의 위치를 예측한 후 파이프라인을 통해 의미역 결정 모델을 학습시킨다. ALBERT 사전학습 모델을 통해 입력 토큰의 표현을 얻은 후, 논항의 위치에 대응되는 표현을 따로 추상화하기 위한 계층형 트랜스포머 인코더 레이어 구조를 추가했다. 실험결과 Korean Propbank 데이터에 대해 F1 85.59의 성능을 보였다.

  • PDF

Korean Dependency Parsing using Pretrained Language Model and Specific-Abstraction Encoder (사전 학습 모델과 Specific-Abstraction 인코더를 사용한 한국어 의존 구문 분석)

  • Kim, Bongsu;Whang, Taesun;Kim, Jungwook;Lee, Saebyeok
    • Annual Conference on Human and Language Technology
    • /
    • 2020.10a
    • /
    • pp.98-102
    • /
    • 2020
  • 의존 구문 분석은 입력된 문장 내의 어절 간의 의존 관계를 예측하기 위한 자연어처리 태스크이다. 최근에는 BERT와 같은 사전학습 모델기반의 의존 구문 분석 모델이 높은 성능을 보이고 있다. 본 논문에서는 추가적인 성능 개선을 위해 ALBERT, ELECTRA 언어 모델을 형태소 분석과 BPE를 적용해 학습한 후, 인코딩 과정에 사용하였다. 또한 의존소 어절과 지배소 어절의 특징을 specific하게 추상화 하기 위해 두 개의 트랜스포머 인코더 스택을 추가한 의존 구문 분석 모델을 제안한다. 실험결과 제안한 모델이 세종 코퍼스에 대해 UAS 94.77 LAS 94.06의 성능을 보였다.

  • PDF

Case Report of Two Stroke Patients with Hemi-spatial Neglect Treated with Traditional Korean Medicine (중풍 환자에게 발생한 편측 무시 한방 치험 2례 보고)

  • Kim, Sae-won;Woo, Seong-jin;Shin, Jae-wook;Baek, Kyung-min;Jang, Woo-seok
    • The Journal of Internal Korean Medicine
    • /
    • v.37 no.2
    • /
    • pp.156-165
    • /
    • 2016
  • Objectives: This study reports on two cases of the clinical application of traditional Korean medicine on stroke patients with hemi-spatial neglect.Methods: We applied several traditional Korean medicine treatments and then evaluated the patients’ symptoms with Albert’s test, the clock-drawing test (CDT), the Catherine Bergego scale (CBS), and the numeric rating scale (NRS).Results: The scores of all of the scales showed improvement. In case 1, the Albert’s test score decreased from 2.5% to 0%, the CDT score increased from 6 to 10, and the CBS score decreased from 16 to 11. In case 2, the Albert’s test score decreased from 5% to 2.5%, the CDT score increased from 3 to 6, and the CBS score decreased from 23 to 16.Conclusion: We found that acupuncture, moxibustion, herbal medicine, and physical therapy appeared to be effective treatments for hemi-spatial neglect in these two stroke patients.

Exploratory Big Data Analysis of Albert Camus's La Peste in Post Corona era (포스트 코로나 시대 알베르 카뮈의 『페스트』에 관한 탐색적 빅데이터 분석)

  • MIN, Jinyoung
    • The Journal of the Convergence on Culture Technology
    • /
    • v.7 no.1
    • /
    • pp.432-438
    • /
    • 2021
  • This dissertation's object is to confirm the drastic popularity of La Peste of Albert Camus in Korea post-corona society using big data as the mean of inductive research. Analyzing news articles concerning Camus and investigating word frequency of the book La Peste will affirm the implications La Peste has on current Korea society as the outbreak spreads. As an analysis tool, Bigkinds of Korea Press Foundation and Nuagedemots, the French version of Word Cloud were used. For the past 30 years, Albert Camus has been known in Korea as the writer of L'étranger, but after the epidemic, he earned more reputation with La Peste. Compared to L'étranger that rebelled against the world's absurdity with ennui, La peste emphasizes the importance of resistance accompanied by solidarity. La peste conveys hope by depicting disastrous situations of citizens who confront the plague by organizing a health college. The novel delivers a lot of ethical inspiration to humanity in this exceptional circumstance of COVID-19.