• Proceedings of the Korean Society for Cognitive Science Conference
• /
• 2000.06a
• /
• pp.33-40
• /
• 2000

복합명사의 한 구조는 구성 명사간의 수식관계의 집합이라고 본다. 한 복합 명사에 대하여 가능한 여러 구조 중에서 올바른 구조를 알아 내는 것이 본 논문의 목표이다. 이를 위하여 우리는 최근에 유행하는 통계 기반 분석 기법을 이용한다. 먼저 우리의 복합 명사 분석 asn제에 알맞은 통계 모델을 개발하였다. 이 모델을 이용하면 분석하려는 복합명사의 가능한 분석 구조바다 확률값을 얻게 된다. 그 다음 가능한 구조들 중에서 가장 확률값이 큰 구조를 복합구조로 선택한다. 통계 기반 기법에서 항상 문제가 되는 것이 데이터 부족문제이다. 우리는 이를 해결하기 위해 개념적 계층구조의 하나인 워드넷(WordNet)을 이용한다.

• Journal of the Korea Computer Graphics Society
• /
• v.4 no.1
• /
• pp.11-19
• /
• 1998

Non-manifold topological operations such as Euler and Boolean operations provide a versatile environment for modeling domains. The implementation of these operations raises geometrical issues that need to be addressed to ensure the topological validity of the underlying model, and they uses the glue operation which provides a basic method to modify the topology of non-manifold models when vertices, edges and faces are contacting each other. Topological information such as adjacency relationships should be inferred when gluing non-manifold models. Two methods of reasoning can be employed to find the topological information : topological reasoning and geometrical reasoning. The topological method can infer the adjacency relationships by using stored topological information. On the other hand, the geometrical method can find topological ambiguities by considering the geometrical shape at the local area of gluing when the topological relations were not stored. This paper describes the geometrical reasoning method.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.10a
• /
• pp.994-997
• /
• 2019

영상 기반 상식 추론(VCR) 문제는 기존의 영상 기반 질문-응답(VQA) 문제들과는 달리, 영상에 포함된 사물들 간의 관계 파악과 답변 근거 제시 등 별도의 상식 추론이 요구되는 새로운 지능 문제이다. 본 논문에서는 입력 데이터(영상, 자연어 질문, 응답 리스트)에서 사물들 간의 관계와 맥락 정보를 추출해내는 모듈들 외에, 별도로 ConceptNet과 같은 외부 지식 베이스로부터 관련 상식들을 직접 가져다 GCN 기반의 지식 그래프 임베딩 과정을 거쳐 추가적으로 활용할 수 있는 모듈들을 포함한 새로운 심층 신경망 모델인 KG_VCR을 제안한다. 제안 모델인 KG_VCR의 세부 설계사항들을 소개하고, VCR 벤치마크 데이터 집합을 이용한 다양한 실험들을 통해 제안 모델의 성능을 입증한다.

• The KIPS Transactions:PartB
• /
• v.16B no.3
• /
• pp.247-256
• /
• 2009

Typical pseudo-relevance feedback methods assume the top-retrieved documents are relevant and use these pseudo-relevant documents to expand terms. The initial retrieval set can, however, contain a great deal of noise. In this paper, we present a cluster-based resampling method to select better pseudo-relevant documents based on the relevance model. The main idea is to use document clusters to find dominant documents for the initial retrieval set, and to repeatedly feed the documents to emphasize the core topics of a query. Experimental results on large-scale web TREC collections show significant improvements over the relevance model. For justification of the resampling approach, we examine relevance density of feedback documents. The resampling approach shows higher relevance density than the baseline relevance model on all collections, resulting in better retrieval accuracy in pseudo-relevance feedback. This result indicates that the proposed method is effective for pseudo-relevance feedback.

• Journal of Service Research and Studies
• /
• v.2 no.2
• /
• pp.35-43
• /
• 2012

Data mining is an interest area in all field around us not in any specific areas, which could be used applications in a number of areas heavily. In other words, it is used in the decision-making process, data and correlation analysis in hidden relations, for finding the actionable information and prediction. But some of the data sets contains many missing values in the variables and do not exist a large number of records in the data set. In this paper, missing values are handled in accordance with the model tree algorithm. Cholesterol value is applied for predicting. For the performance analysis, experiments are approached for each treatment. Through this, efficient alternative is presented to apply the missing data.

• The Journal of Society for e-Business Studies
• /
• v.18 no.2
• /
• pp.95-110
• /
• 2013

This paper presents a novel feature engineering technique that can improve the conventional machine learning-based text classification systems. The proposed method extends the initial set of features by using hyperlink relationships in order to effectively categorize hypertext web documents. Web documents are connected to each other through hyperlinks, and in many cases hyperlinks exist among highly related documents. Such hyperlink relationships can be used to enhance the quality of features which consist of classification models. The basic idea of the proposed method is to generate a sort of ed concept feature which consists of a few raw feature words; for this, the method computes the semantic similarity between a target document and its neighbor documents by utilizing hierarchical relationships in the WordNet ontology. In developing classification models, the ed concept features are equated with other raw features, and they can play a great role in developing more accurate classification models. Through the extensive experiments with the Web-KB test collection, we prove that the proposed methods outperform the conventional ones.

• The Journal of the Korea Contents Association
• /
• v.6 no.1
• /
• pp.23-30
• /
• 2006

A clustering method for temporal data takes a model-based approach. This uses automata based model for each cluster. It is necessary to construct global models for a set of data in order to elicit individual models for the cluster. The preparation for building individual models is completed by determining the number of clusters inherent in the data set. In this paper, BIC(Bayesian Information Criterion) approximation is used to determine the number clusters and confirmed its applicability. A search technique to improve efficiency is also suggested by analyzing the relationship between data size and BIC values. A number of experiments have been performed to check its validity using artificially generated data sets. BIC approximation measure has been confirmed that it suggests best number of clusters through experiments provided that the number of data is relatively large.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.377-381
• /
• 2023

재순위화기 연구는 주로 파이프라인 과정 설계, 데이터 증강, 학습 함수 개선, 혹은 대규모 언어 모델의 지식 활용 등에 집중되어있다. 이러한 연구들은 좋은 성능 상승을 이끌어주었지만 실제 적용이 힘들 뿐만 아니라 학습 비용이 크게 발생한다는 한계점을 가지고 있다. 더 나아가 주어진 데이터 집합만을 활용해서는 보다 더 세부적인 학습 신호를 주기 어렵다는 단점 또한 존재한다. 최근 자연어처리 분야의 연구에서는 피드백을 인위적으로 생성하여 반영하여 모델 성능 상승을 이끄는 연구가 제안되었다. 본 연구는, 이러한 연구를 바탕으로 질의와 문서 간의 함의 관계 점수를 피드백으로 사용 및 재순위화기 모델로의 반영을 제안한다. 재순위화기 모델에 대해 피드백을 반영하는것은 그렇지 않은 모델 대비하여 성능 상승을 이끌며 피드백 반영이 더 좋은 표상 도출에 도움이 됨을 확인할 수 있다.

• Journal of KIISE:Databases
• /
• v.31 no.3
• /
• pp.205-218
• /
• 2004

As the development of database technology for managing spatiotemporal data, new types of spatiotemporal application services that need the spatiotemporal knowledge discovery from the large volume of spatiotemporal data are emerging. In this paper, a new 3-layered discovery framework for the development of spatiotemporal knowledge discovery techniques is proposed. The framework supports the foundation model in order not only to define spatiotemporal knowledge discovery problem but also to represent the definition of spatiotemporal knowledge and their relationships. Also the components of spatiotemporal knowledge discovery system and its implementation model are proposed. The discovery framework proposed in this paper satisfies the requirement of the development of new types of spatiotemporal knowledge discovery techniques. The proposed framework can support the representation model of each element and relationships between objects of the spatiotemporal data set, information and knowledge. Hence in designing of the new types of knowledge discovery such as spatiotemporal moving pattern, the proposed framework can not only formalize but also simplify the discovery problems.

• Journal of Educational Research in Mathematics
• /
• v.18 no.3
• /
• pp.373-389
• /
• 2008

In this paper, we designed a teaching unit for the learning mathematization of secondary pre-service teachers through exploring the relationship between partition models and generalized fibonacci sequences. We first suggested some problems which guide pre-service teachers to make phainomenon for organizing nooumenon. Pre-service teachers should find patterns from partitions for various partition models by solving the problems and also form formulas front the patterns. A series of these processes organize nooumenon. Futhermore they should relate the formulas to generalized fibonacci sequences. Finding these relationships is a new mathematical material. Based on developing these mathematical materials, pre-service teachers can be experienced mathematising as real practices.