• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.143-148
• /
• 2023

본 논문에서는 금융 도메인 특화 사전학습 언어모델인 KF-DeBERTa(Korean Finance DeBERTa)를 제안한다. KF-DeBERTa는 대규모의 금융 말뭉치를 기반으로 학습하였으며, Transformer 아키텍처와 DeBERTa의 특징을 기반으로 구성되었다. 범용 및 금융 도메인에 대한 평가에서 KF-DeBERTa는 기존 언어모델들에 비해 상당히 높은 성능을 보였다. 특히, 금융 도메인에서의 성능은 매우 두드러졌으며, 범용 도메인에서도 다른 모델들을 상회하는 성능을 나타냈다. KF-DeBERTa는 모델 크기 대비 높은 성능 효율성을 보여주었고, 앞으로 금융 도메인에서의 활용도가 기대된다.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.704-706
• /
• 2023

이 논문에서는 과학기술분야 특화 한국어 사전학습 언어모델인 KorSciDeBERTa를 소개한다. DeBERTa Base 모델을 기반으로 약 146GB의 한국어 논문, 특허 및 보고서 등을 학습하였으며 모델의 총 파라미터의 수는 180M이다. 논문의 연구분야 분류 태스크로 성능을 평가하여 사전학습모델의 유용성을 평가하였다. 구축된 사전학습 언어모델은 한국어 과학기술 분야의 여러 자연어처리 태스크의 성능향상에 활용될 것으로 기대된다.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.5
• /
• pp.221-226
• /
• 2024

In information retrieval, queries come in various types, ranging from abstract queries to those containing specific keywords, making it a challenging task to accurately produce results according to user demands. Additionally, search systems must handle queries encompassing various elements such as typos, multilingualism, and codes. Reranking is performed through training suitable documents for queries using DeBERTa, a deep learning model that has shown high performance in recent research. To evaluate the effectiveness of the proposed method, experiments were conducted using the test collection of the Product Search Track at the TREC 2023 international information retrieval evaluation competition. In the comparison of NDCG performance measurements regarding the experimental results, the proposed method showed a 10.48% improvement over BM25, a basic information retrieval model, in terms of search through query error handling, provisional relevance feedback-based product title-based query expansion, and reranking according to query types, achieving a score of 0.7810.

• ETRI Journal
• /
• v.28 no.2
• /
• pp.257-259
• /
• 2006

In this letter, we will evaluate the performance degradation of a 40 km high-speed (40 Gb/s) optical system, induced by optical fiber variations of the chromatic dispersion induced by temperature changes. The chromatic dispersion temperature sensitivity will be estimated based on the signal quality parameters.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.206-208
• /
• 2022

As the big data industry has recently developed significantly, interest in privacy violations caused by personal information leakage has increased. There have been attempts to automate this through named entity recognition in natural language processing. In this paper, named entity recognition data is constructed semi-automatically by identifying sentences with de-identification information from de-identification information in Korean Wikipedia. This can reduce the cost of learning about information that is not subject to de-identification compared to using general named entity recognition data. In addition, it has the advantage of minimizing additional systems based on rules and statistics to classify de-identification information in the output. The named entity recognition data proposed in this paper is classified into twelve categories. There are included de-identification information, such as medical records and family relationships. In the experiment using the generated dataset, KoELECTRA showed performance of 0.87796 and RoBERTa of 0.88.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.07a
• /
• pp.529-532
• /
• 2023

본 논문은 소셜 네트워크 이용자의 텍스트 데이터를 대상으로, 트랜스포머 계열의 언어모델을 전이학습해 이용자의 MBTI 성격 유형을 분류한 국내 첫 연구이다. Kaggle MBTI Dataset을 대상으로 RoBERTa Distill, DeBERTa-V3 등의 사전 학습모델로 전이학습을 해, MBTI E/I, N/S, T/F, J/P 네 유형에 대한 분류의 평균 정확도는 87.9181, 평균 F-1 Score는 87.58를 도출했다. 해외 연구의 State-of-the-art보다 네 유형에 대한 F1-Score 표준편차를 50.1% 낮춰, 유형별 더 고른 분류 성과를 보였다. 또, Twitter, Reddit과 같은 글로벌 소셜 네트워크 서비스의 텍스트 데이터를 추가로 분류, 트랜스포머 기반의 MBTI 분류 방법론을 확장했다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.1045-1046
• /
• 2023

본 논문은 디지털 소외계층과 사회적 약자를 고려한 보이스피싱 예방 솔루션을 제안한다. 통화 내용을 AWS Transcribe를 활용한 STT와 NLP 알고리즘을 사용해 실시간으로 보이스피싱 위험도를 파악하고 결과를 사용자에게 전달하도록 한다. NLP 알고리즘은 KoBIGBIRD와 DeBERTa 모델 각각을 커스터마이즈하여 보이스피싱 탐지에 적절하게 파인튜닝 했다. 이후, 성능과 인퍼런스를 비교하여 더 좋은 성능을 보인 KoBIGBIRD 모델로 보이스피싱 탐지를 수행한다.

• Publications of The Korean Astronomical Society
• /
• v.32 no.1
• /
• pp.239-244
• /
• 2017

How does the presence of an AGN influence the total SFR estimates of galaxies and change their distribution with respect to the Galaxy Main Sequence? To contribute to solving this question, we study a sample of 1133 sources detected in the North Ecliptic Pole field (NEP) by AKARI and Herschel. We create a multi-wavelength dataset for these galaxies and we fit their multi-wavelength Spectral Energy Distribution (SED) using the whole spectral regime (from 0.1 to $500{\mu}m$). We perform the fit using three procedures: LePhare and two optimised codes for identifying AGN tracers from the SED analysis. In this work we present an overview of the comparison between the estimates of the Infrared bolometric luminosities (between 8 and $1000{\mu}m$) and the AGN fractions obtained exploiting these different procedures. In particular, by estimating the AGN contribution in four different wavelength ranges ($5-40{\mu}m$, $10-20{\mu}m$, $20-40{\mu}m$ and $8-1000{\mu}m$) we show how the presence of an AGN affects the PAH emission by suppressing the ratio $\frac{L_{8{\mu}m}}{L_{4.5{\mu}m}}$ as a function of the considered wavelength range.

• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.191-206
• /
• 2022

Recently, it is a de-facto approach to utilize a pre-trained language model(PLM) to achieve the state-of-the-art performance for various natural language tasks(called downstream tasks) such as sentiment analysis and question answering. However, similar to any other machine learning method, PLM tends to depend on the data distribution seen during the training phase and shows worse performance on the unseen (Out-of-Distribution) domain. Due to the aforementioned reason, there have been many efforts to develop domain-specified PLM for various fields such as medical and legal industries. In this paper, we discuss the training of a finance domain-specified PLM for the Korean language and its applications. Our finance domain-specified PLM, KB-BERT, is trained on a carefully curated financial corpus that includes domain-specific documents such as financial reports. We provide extensive performance evaluation results on three natural language tasks, topic classification, sentiment analysis, and question answering. Compared to the state-of-the-art Korean PLM models such as KoELECTRA and KLUE-RoBERTa, KB-BERT shows comparable performance on general datasets based on common corpora like Wikipedia and news articles. Moreover, KB-BERT outperforms compared models on finance domain datasets that require finance-specific knowledge to solve given problems.