• Korean Journal of Cognitive Science
• /
• v.16 no.2
• /
• pp.69-92
• /
• 2005

A brief survey of linguistic studies on the nature of contrastive information in discourse was Presented first, and an attempt was also made to incorporate the Linguistic theories and concepts about contrast in discourse into a psychological framework. A tentative model of processing of contrastive information in discourse was Proposed, and eight experimental studies on the effects of contrastive information on comprehension and memory of short and ions discourses were reviewed. Experimental results showed that contrastive sentences took more time to process at encoding, and yet were recognized faster and cued-recalled in greater amount than noncontrastive sentences. It was also found that levels of contrast in the discourse structure have some effects on encoding time. It was further found that the sentence immediately following the contrastive sentence was processed slowly regardless of whether it does or does not resolve the contrast. The implications of the results of empirical studies were discussed in relation to developing a research framework that integrate coherence studies and contrast studies urns the two disciplines of linguistics and cognitive psychology.

• Proceedings of the Korean Society for Cognitive Science Conference
• /
• 2006.06a
• /
• pp.125-130
• /
• 2006

원활한 의사소통을 위해서는 문장들을 연결하여 흐름을 조직하고 말로 산출하고 전체적인 의미를 파악할 수 있어야 한다. 이야기는 이러한 문장들이 연결되어있는 것으로, 종속적이거나 나열적인 이야기 특성은 의사소통장애인의 이야기 이해와 산출의 수행에 영향을 미칠 수 있다. 본 연구에서는 이야기 특성에 따른 유창성 실어증환자의 이야기 이해 및 산출의 능력을 알아보고, 이해과제 수행이 산출과제에 미치는 영향을 분석해보았다. 이야기 종류로는 시간적 나열 이야기와 인과적 관계 이야기, 유머가 있는 이야기를 사용하였으며, 사실적 정보, 텍스트 추론, 빠진 정보추론 등 세 가지의 이해과제를 통하여 이해 능력을 측정하였다. 산출능력은 이해과제 전과 후에 CIU 비율로 질적인 측면을 측정하고, 분당어절 수를 이용하여 양적인 측면을 분석하였다. 그 결과 이해측면은 세 가지 이야기 모두 사실적 정보에 대한 이해 능력이 상대적으로 좋았으며, 오류의 형태는 추론오류가 가장 많이 나타났다. 산출에서는 인과적 관계이야기에서의 CIU 비율이 가장 높았고, 이해과제 전, 후의 차이를 비교한 결과, CIU 비율은 변화하지 않았으나, 분당 어절수에서는 증가하고 있음을 보여주었다. 이야기의 종류에 따라서 유창성 실어증화자의 산출과제의 수행수준은 다르게 나타났다. 그리고 이해과제의 수행이 산출과제에서 양적인 증가는 가져왔으나 질적인 수준에는 아무런 영향을 미치지 않았다.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.11
• /
• pp.105-113
• /
• 2014

In this paper, we propose an OWL DL mapping rules for construction legal ontology based on the analyzed relationship between the structural features and elements of the statute. The mapping rule to be proposed is the method building the structure of the domestic statute, unique attribute of the statute, and reference relation between laws with TBox, and the legal sentence is analyzed, and the pattern type of the sentence is selected. It expresses with ABox. The proposed mapping rule is transformed to the information in which the computer can process the domestic legal document. It is usable for the legal knowledge base.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.461-464
• /
• 2023

최근 GPT-3 와 LLaMa 같은 생성형 거대 언어모델을 활용한 서비스가 공개되었고, 실제로 많은 사람들이 사용하고 있다. 해당 모델들은 사용자들의 다양한 질문에 대해 유창한 답변을 한다는 이유로 주목받고 있다. 하지만 LLMs 의 답변에는 종종 Inconsistent content 와 non-factual statement 가 존재하며, 이는 사용자들로 하여금 잘못된 정보의 전파 등의 문제를 야기할 수 있다. 이에 논문에서는 동일한 질문에 대한 LLM 의 답변 샘플과 외부 지식을 활용한 Hallucination Detection 방법을 제안한다. 제안한 방법은 동일한 질문에 대한 LLM 의 답변들을 이용해 일관성 점수(Consistency score)를 계산한다. 거기에 외부 지식을 이용한 사실검증을 통해 사실성 점수(Factuality score)를 계산한다. 계산된 일관성 점수와 사실성 점수를 활용하여 문장 수준의 Hallucination Detection 을 가능하게 했다. 실험에는 GPT-3 를 이용하여 WikiBio dataset 에 있는 인물에 대한 passage 를 생성한 데이터셋을 사용하였으며, 우리는 해당 방법을 통해 문장 수준에서의 Hallucination Detection 성능이 baseline 보다 AUC-PR scores 에서 향상됨을 보였다.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.27 no.2
• /
• pp.167-176
• /
• 2024

Lip-reading is the task of inferring the speaker's utterance from silent video based on learning of lip movements. It is very challenging due to the inherent ambiguities present in the lip movement such as different characters that produce the same lip appearances. Recent advances in deep learning models such as Transformer and Temporal Convolutional Network have led to improve the performance of lip-reading. However, most previous works deal with English lip-reading which has limitations in directly applying to Korean lip-reading, and moreover, there is no a large scale Korean lip-reading dataset. In this paper, we introduce the first large-scale Korean lip-reading dataset with more than 120 k utterances collected from TV broadcasts containing news, documentary and drama. We also present a preprocessing method which uniformly extracts a facial region of interest and propose a transformer-based model based on grapheme unit for sentence-level Korean lip-reading. We demonstrate that our dataset and model are appropriate for Korean lip-reading through statistics of the dataset and experimental results.

• Korean Journal of Cognitive Science
• /
• v.2 no.2
• /
• pp.279-307
• /
• 1990

The present research inverstigates the nature of contextual effects on pronoun reference resolution during text comprehesion.Through three experiments,this research examines how various contextuall informations influence on reference resolution and interact with syntactic variables.In experiment 1,the local context was controlled by biasing the pronoun-sentence context towards a certain preceding referent.The lexical decision time and the forced choice time for the correct referent were measured.The results showed that the local contexts have clear effect on reference resolution.The effects of syntactic ambiguity were also observed though the local context was biased towards a certain referent noun.In experiment 2,the global context effect was examined by introducing the text-thematic context in a preceding sentence while keeping the following pronoun-sentence context neutral.The results showed that the global thematic context bias towards a subject or object in a preceding sentence entails a faster response time than the thematically neutral context.In experiment 3,another aspects of context effects were inverstigated by manipulating the consistency of the preceding thematic context with the following pronoun-sentence context.The results showed that the lexical decision responses and forced referent choice responses were faster when the prethematic context and the post-anaphoric context match than when they mismatch.In sum,the overall results of three experiments of this research indicates that context has a clear effect on pronoun reference resolution during text comprehension.

• The Journal of the Korea Contents Association
• /
• v.20 no.10
• /
• pp.25-34
• /
• 2020

Community question and answering service is one of the main sources of information and knowledge in the Web. The quality of information in question and answer documents is determined by the clarity of the question and the relevance of the answers, and the readability of a document is a key factor for evaluating the quality. This study is to measure the quality of documents used in community question and answering service. For this purpose, we compare the frequency of occurrence by vocabulary level used in community documents and measure the readability index of documents by institution of author. To measure the readability index, we used the Dale-Chall formula which is calculated by vocabulary level and sentence length. The results show that the vocabulary used in the answers is more difficult than in the questions and the sentence length is longer. The gap in readability between questions and answers is also found by writing institution. The results of this study can be used as basic data for improving online counseling services.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.8
• /
• pp.694-702
• /
• 2003

The word intelligibility test for dysarthric speakers was designed to examine phonetic contrasts that are likely (1) to be sensitive to intelligibility impairment and (2) to contribute significantly to speech intelligibility. These phonetically contrasting word pairs were tested and proved to be reliable and to be valid, The results showed that in Korean dysarthric patients, the percentage of error in final position contrast was higher than in any other position. Unlike the results of previous studies, the initial-position contrasts were crucial in predicting the overall intelligibility among Korean patients.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.243-248
• /
• 2020

인공 신경망을 통한 심층 학습 모델의 발전은 컴퓨터 비전, 자연언어 이해 문제들에서 인간을 뛰어넘는 성능을 보이고 있다. 특히 트랜스포머[1] 기반의 사전 학습 모델은 질의응답, 대화문과 같은 자연언어 이해 문제에서 최근 높은 성능을 보이고 있다. 하지만 트랜스포머 기반의 모델과 같은 심층 학습 모델의 급격한 발전 양상에 비해, 이의 동작 방식은 상대적으로 잘 알려져 있지 않다. 인공 신경망을 통한 심층 학습 모델을 해석하는 방법으로 모델의 예측 값과 실제 값이 얼마나 일치하는지를 측정하는 모델의 보정(Calibration)이 있다. 본 연구는 한국어 기반의 심층학습 모델의 해석을 위해 모델의 보정을 수행하였다. 그리고 사전 학습된 한국어 언어 모델이 문장이 내포하는 애매성을 잘 파악하는지의 여부를 확인하고, 완화 기법들을 적용하여 문장의 애매성을 확신 수준을 통해 정량적으로 출력할 수 있도록 하였다. 또한 한국어의 문법적 특징으로 인한 문장의 의미 변화를 모델 보정 관점에서 평가하여 한국어의 문법적 특징을 심층학습 언어 모델이 잘 이해하고 있는지를 정량적으로 확인하였다.

• Journal of the Korean Society for information Management
• /
• v.22 no.3 s.57
• /
• pp.213-232
• /
• 2005

The purpose of this study is to determine factors affecting text difficulty and to model objective formulas which measure readability scores. Some readability-related factors such as total number of letters, total number of syllables, total number of unique syllables, total number of sentences and total number of paragraphs were found through correlation analysis. Some regression equations with these factors as their variables were produced through regression analysis. A model estimating readability score from total number of unique syllables was a good formula, while a model with two factors, total number of unique syllables and new syllable occurrence ratio, was a better enhanced one. The readability score represents detailed level so we can recommend students read texts corresponding to their reading levels.