• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.468-473
• /
• 2023

본 논문은 사전 학습 언어 모델의 특성인 이방성과 문맥화에 주목하여 이에 대한 분석 실험과 한국어 언어 모델만의 새로운 관점을 제안한다. 최근 진행된 영어 언어 모델 분석 연구에서 영감을 받아, 한국어 언어 모델에서도 대조학습을 통한 이방성과 문맥화의 변화를 보고하였으며, 다양한 모델에 대하여 토큰들을 문맥화 정도에 따라 분류하였다. 또한, 한국어의 언어학적 특성을 고려하여, 허위 문맥화를 완화할 수 있는 토큰을 문맥 중심어로, 문맥 중심어의 임베딩을 모방하는 토큰을 문맥 기능어로 분류하는 기준을 제안하였다. 간단한 적대적 데이터 증강 실험을 통하여 제안하는 분류 기준의 가능성을 확인하였으며, 본 논문이 향후 평가 벤치마크 및 데이터셋 제작, 나아가 한국어를 위한 강건한 학습 방법론에 기여하길 바란다.

• Annual Conference on Human and Language Technology
• /
• 2009.10a
• /
• pp.67-72
• /
• 2009

ERP의 일종인 MMN(Mismatch Negativity)은 언어의 청각 인지정보 처리과정(central auditory processing)을 규명하는 데 유용한 수단으로 이용되어 왔다. 그런데, 최근의 연구들은 이러한 MMN이 청각 자극뿐만 아니라 시각 자극에 의해서도 검출될 수 있음을 밝혀냈다. 본 연구는 이러한 시각적 MMN을 이용하여 뇌에서 이루어지는 한국어 화자의 무의식적인 한국어 문자 정보처리과정을 규명하려고 시도하였다. 본 연구에서는 한국어의 글말 최소쌍 '므'/'모'와 '므'/'무', 이에 대응되는 비언어자극 '+ㅡ'/'+ㅗ'와 '+ㅡ'/'+ㅜ'(+표시의 아래에 모음을 붙여서 만든 인공문자, 그림1 참고)를 수동적(passive) Oddball paradigm으로 제시하고 언어 자극에 대한 EEG를 비언어자극과 비교 하에 측정, 분석하였다. 본 연구의 결과, 언어자극과 비언어자극 모두에서 시각적 MMN이 검출되었다. 하지만, 언어자극의 시각적 MMN이 비언어자극의 시각적 MMN보다 높게 나타남을 확인하였다. 이는 한국어 모국어화자들이 무의식적인 인지과정에서 언어자극이 갖는 물리적인 시각 정보뿐만 아니라 한국어 문자의 언어적 정보도 함께 처리하고 있음을 보여주는 것이다. 본 연구의 결과들은 한국어 글말의 무의식적인 인지처리과정을 밝혀주는 한편, 한국어 문자가 인지과학에서 갖는 중요한 지위를 보여줄 수 있을 것으로 기대된다.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.238-244
• /
• 2022

감정 분석은 텍스트에 표현된 인간의 감정을 인식하여 다양한 감정 유형으로 분류하는 것이다. 섬세한 인간의 감정을 보다 정확히 분류하기 위해서는 감정 유형의 분류가 무엇보다 중요하다. 본 연구에서는 사전 학습 언어 모델을 활용하여 우리말샘의 감정 어휘와 용례를 바탕으로 기쁨, 슬픔, 공포, 분노, 혐오, 놀람, 흥미, 지루함, 통증의 감정 유형으로 분류된 감정 말뭉치를 구축하였다. 감정 말뭉치를 구축한 후 성능 평가를 위해 대표적인 트랜스포머 기반 사전 학습 모델 중 RoBERTa, MultiDistilBert, MultiBert, KcBert, KcELECTRA. KoELECTRA를 활용하여 보다 넓은 범위에서 객관적으로 모델 간의 성능을 평가하고 각 감정 유형별 정확도를 바탕으로 감정 유형의 특성을 알아보았다. 그 결과 각 모델의 학습 구조가 다중 분류 말뭉치에 어떤 영향을 주는지 구체적으로 파악할 수 있었으며, ELECTRA가 상대적으로 우수한 성능을 보여주고 있음을 확인하였다. 또한 감정 유형별 성능을 비교를 통해 다양한 감정 유형 중 기쁨, 슬픔, 공포에 대한 성능이 우수하다는 것을 알 수 있었다.

• The KIPS Transactions:PartD
• /
• v.11D no.2
• /
• pp.415-422
• /
• 2004

Automation systems improve the productivity of works which relate to product design, facilities management, fault processing and quality evaluation. In these systems, the description language for monitoring and control process is called new control language. These are five flow control languages : IL, ST, FBD, SFC and LD. IL and ST are based on left form. FBD, SFC and LD are based on graphic form. Generally, a software which monitors and controls a system is allowed to use just one flow control language. It is impossible to use more than two languages for simulation in the same system environment. In this paper, we analyzed the characteristics of flow control languages and the process of programming in the legacy system. In addition, for the Integrated processing of languages, we propose Extended ST based on the high-level ST language. Based on this research, we implement a graphical language editor and EST-IL convertor. The graphical language editor makes sequence rules, and converts graphical language into EST. EST-IL convertor has a function to convert EST into IL which is similar to assembly language. As the result of this paper, we present a scheme which integrates all the flow control language processing based on IL.

• Korean Journal of Cognitive Science
• /
• v.10 no.2
• /
• pp.35-43
• /
• 1999

This paper aims to research the language changes in modern Korean and its effect to the language processing systems. In modern Korean. the syntactic constructions l like [Adnominal Ending + Bound Noun ( + Postposition)] are changing into the morphological constructions, and some of these constructions are reflected in the written language. For example. the syntactic construction [Ad nominal Ending + '-de (Bound N Noun)' (+ Postposition) ) co-exists with the mixed form '-neunde' and [Adnominal Ending + 'geot' (Bound Noun) + '-j' (Postposition)) does with ' neunge'. These constructions are used frequently in the spoken language. As like other verbal endings, these forms also participate in the construction of the complex sentence, and these forms have its own case function fused into themselves So, the analytic approach to these forms can make great effect on the automatic morphological analysis systems. automatic tagging systems. and the syntactic analysis systems. So. in the design phase of a language processing systems, the language change phenomena like these must be taken l into consideration.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2017.07a
• /
• pp.73-74
• /
• 2017

4차 산업혁명의 핵심 중 하나인 소프트웨어 교육에서 프로그래밍 교육은 필수적이다. 다양한 프로그래밍 언어 중 프로그래밍 입문자를 위해 적합한 언어를 선택하는 것은 매우 중요하다. 본 논문에서는 프로그래밍 입문을 위해 적합한 언어를 선택하기 위한 가이드를 제공하기 위해 2017학년도 1학기에 프로그래밍 입문 관련 교과목을 수강한 학생들을 대상으로 프로그래밍 언어에 대한 선호도를 조사하고 분석하여 언어 선택을 위한 가이드를 제안한다.

• Journal of Intelligence and Information Systems
• /
• v.29 no.3
• /
• pp.401-418
• /
• 2023

Recently, the socio-economic interest in Large Language Models (LLMs) has been growing due to the emergence of ChatGPT. As a type of generative AI, LLMs have reached the level of script creation. In this regard, it is important to address the issue of discrimination (sexism, racism, religious discrimination, ageism, etc.) in the performing arts in general or in specific performing arts works or organizations in a large language model that will be widely used by the general public and professionals. However, there has not yet been a full-scale investigation and discussion on the issue of discrimination in the performing arts in large-scale language models. Therefore, the purpose of this study is to textually analyze the perceptions of discrimination issues in the performing arts from LMMs and to derive implications for the performing arts field and the development of LMMs. First, BBQ (Bias Benchmark for QA) questions and measures for nine discrimination issues were used to measure the sensitivity to discrimination of the giant language models, and the answers derived from the representative giant language models were verified by performing arts experts to see if there were any parts of the giant language models' misperceptions, and then the giant language models' perceptions of the ethics of discriminatory views in the performing arts field were analyzed through the content analysis method. As a result of the analysis, implications for the performing arts field and points to be noted in the development of large-scale linguistic models were derived and discussed.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.9 no.9
• /
• pp.797-808
• /
• 2019

The purpose of this study is to collect the psychological responses that occur while watching TV home shopping ads in verbal form, and explore the possibility of measuring the effectiveness of TV home shopping ads using linguistic analysis. The psychological responses during watching positive and negative ads of participants(40 housewives and female college students) were collected in a linguistic form using a think aloud and self-report measurement. It was analyzed by KLIWC, a Korean language analysis program. As a result of the analysis, there was a difference in psychosocial variables as well as linguistic variables between positive and negative evaluation ads. Also, various variables of KLIWC were correlated with the variables of advertising effectiveness (purchase stimulus, ad attitude, product attitude, purchase intention) and advertising response variables. This suggests the possibility of constructing a psychological response profile and measurement of advertising effectiveness using language analysis.

• Journal of Convergence for Information Technology
• /
• v.11 no.3
• /
• pp.147-158
• /
• 2021

The purpose of this study is to perform of composition of young children teacher's verbal interactions and problem solving strategies by semantic network analysis. The data was collected the narrative of teacher in young children's environmental education. Additionally, this study used categories as a unit of analysis in order to find out structural characteristics. The results of this study were as follows. First, teacher's verbal interactions, providing information was the most cental, high influence appeared along with integration. Second, teacher's problem solving strategies, suggestion was the most cental, high influence appeared along with divergent thinking. Third, teacher's verbal interactions and problem solving strategies, suggestion was the most cental, providing information and divergent thinking appeared a strong connection. Besides, suggestion and decision were mediated by concern of problem, assessment appeared connection. Through these results from this study it is suggested that in order to interact and support problem solving process in young children's environmental education, teacher education to respond and understand contextually with young children is necessary.

• Journal of The Korean Association of Information Education
• /
• v.16 no.4
• /
• pp.475-482
• /
• 2012

A class is a complex and interactive process between a teacher and students. This paper describes a software design and implementation of the tool easily to analyze and quantize teaching ability. Our proposed software is implemented based on Flanders linguistic method. The software can easily represent the linguistic analysis results by abstracted graphic drawings. We carry out the performance test of our implemented software targeting incumbent teachers. In the experimental, we could demonstrate the effectiveness and practicality of our implemented software.