• Journal of the Korea Society of Computer and Information
• /
• v.23 no.6
• /
• pp.1-7
• /
• 2018

A multi-label classification is to find multiple labels associated with the input pattern. Multi-label classification can be achieved by extending conventional single-label classification. Common extension techniques are known as Binary relevance, Label powerset, and Classifier chains. However, most of the extended multi-label naive bayes classifier has not been able to accurately estimate posterior probabilities because it does not reflect the label dependency. And the remaining extended multi-label naive bayes classifier has a problem that it is unstable to estimate posterior probability according to the label selection order. To estimate posterior probability well, we propose a new posterior probability estimation method that reflects the probability between all labels and labels efficiently. The proposed method reflects the correlation between labels. And we have confirmed through experiments that the extended multi-label naive bayes classifier using the proposed method has higher accuracy then the existing multi-label naive bayes classifiers.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2018.01a
• /
• pp.147-148
• /
• 2018

다중 레이블 분류 문제는 다중 레이블 데이터를 입력받았을 때 연관된 다수의 레이블을 추측하는 문제이다. 본 논문에서는 다중 레이블 분류 문제의 기법 중 하나인 나이브 베이지안 분류기에 레이블 의존성을 계산하여 결과에 반영한 결과 다중 레이블 분류 문제의 성능이 개선됨을 확인하였다.

• Journal of KIISE:Software and Applications
• /
• v.31 no.2
• /
• pp.235-244
• /
• 2004

We propose a format of a binary phrase structure grammar with composite labels. The grammar adopts binary rules so that the dependency between two sub-trees can be represented in the label of the tree. The label of a tree is composed of two attributes, each of which is extracted from each sub-tree so that it can represent the compositional information of the tree. The composite label is generated from part-of-speech tags using an automatic labeling algorithm. Since the proposed rule description scheme is binary and uses only part-of-speech information, it can readily be used in dependency grammar and be applied to other languages as well. In the best-1 context-free cross validation on 31,080 tree-tagged corpus, the labeled precision is 79.30%, which outperforms phrase structure grammar and dependency grammar by 5% and by 4%, respectively. It shows that the proposed rule description scheme is effective for parsing Korean.

• Journal of KIISE
• /
• v.44 no.8
• /
• pp.822-831
• /
• 2017

In this paper, we propose a Korean dependency parsing model using multi-task learning based pointer networks. Multi-task learning is a method that can be used to improve the performance by learning two or more problems at the same time. In this paper, we perform dependency parsing by using pointer networks based on this method and simultaneously obtaining the dependency relation and dependency label information of the words. We define five input criteria to perform pointer networks based on multi-task learning of morpheme in dependency parsing of a word. We apply a fine-tuning method to further improve the performance of the dependency parsing proposed in this paper. The results of our experiment show that the proposed model has better UAS 91.79% and LAS 89.48% than conventional Korean dependency parsing.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.9
• /
• pp.4665-4683
• /
• 2019

Previous methods build image annotation model by leveraging three basic dependencies: relations between image and label (image/label), between images (image/image) and between labels (label/label). Even though plenty of researches show that multiple dependencies can work jointly to improve annotation performance, different dependencies actually do not "work jointly" in their diagram, whose performance is largely depending on the result predicted by image/label section. To address this problem, we propose the adaptive attention annotation model (AAAM) to associate these dependencies with the prediction path, which is composed of a series of labels (tags) in the order they are detected. In particular, we optimize the prediction path by detecting the relevant labels from the easy-to-detect to the hard-to-detect, which are found using Binary Cross-Entropy (BCE) and Triplet Margin (TM) losses, respectively. Besides, in order to capture the inforamtion of each label, instead of explicitly extracting regional featutres, we propose the self-attention machanism to implicitly enhance the relevant region and restrain those irrelevant. To validate the effective of the model, we conduct experiments on three well-known public datasets, COCO 2014, IAPR TC-12 and NUSWIDE, and achieve better performance than the state-of-the-art methods.

• Journal of KIISE:Software and Applications
• /
• v.35 no.12
• /
• pp.764-773
• /
• 2008

The web ontology language(OWL) has become a W3C recommendation to publish and share ontologies on the semantic web. In order to check the satisfiablity of concepts in OWL ontology, OWL reasoners have been introduced. But most reasoners simply report check results without providing a justification for any arbitrary entailment of unsatisfiable concept in OWL ontologies. In this paper, we propose dependency label based causing inconsistency axiom (CIA) detection for debugging unsatisfiable concepts in ontology. CIA is a set of axioms to occur unsatisfiable concepts. In order to detect CIA, we need to find axiom to cause inconsistency in ontology. If precise CIA is gave to ontology building tools, these ontology tools display CIA to debug unsatisfiable concepts as suitable presentation format. Our work focuses on two key aspects. First, when a inconsistency ontology is given, it detect axioms to occur unsatisfiable and identify the root of them. Second, when particular unsatisfiable concepts in an ontology are detected, it extracts them and presents to ontology designers. Therefore we introduce a tableau-based decision procedure and propose an improved method which is dependency label based causing inconsistency axiom detection. Our results are applicable to the very expressive logic SHOIN that is the basis of the Web Ontology Language.

• 한국어정보학회:학술대회논문집
• /
• 2016.10a
• /
• pp.14-19
• /
• 2016

본 논문은 문장에서의 어절 간 의존관계가 성립될 때 의존소와 지배소가 어떠한 관계를 가지는지 의존 관계명을 부착하는 모델을 제안한다. 국내에서 한국어 의존구문분석에 관한 연구가 활발히 진행되고 있지만 의존 관계만을 결과로 제시하고 의존 관계명을 제공하지 않는 경우가 많았다. 따라서 본 논문에서는 의존경로(Dependency Path)와 음절의 의존 관계명 분포를 반영하는 음절 임베딩를 이용한 의존 관계명 부착모델을 제안한다. 문장에서 나올 수 있는 최적의 입력 열인 의존 경로(Dependency Path)를 순차 레이블링에서 좋은 성능을 나타내고 있는 bidirectional LSTM-CRFs의 입력 값으로 사용하여 의존 관계명을 결정한다. 제안된 기법은 자질에 대한 많은 노력 없이 의존 경로에 따라 어절 및 음절 단어표상(word embedding)만을 사용하여 순차적으로 의존 관계명을 부착한다. 의존 경로를 사용하지 않고 전체 문장의 어절 순서를 바탕으로 자질을 추출하여 CRFs로 분석한 기존 모델보다 의존 경로를 사용했을 때 4.1%p의 성능향상을 얻었으며, 의존 관계명 분포를 반영하는 음절 임베딩을 사용한 bidirectional LSTM-CRFs는 의존 관계명 부착에 최고의 성능인 96.01%(5.21%p 개선)를 내었다.

• Annual Conference on Human and Language Technology
• /
• 2016.10a
• /
• pp.14-19
• /
• 2016

본 논문은 문장에서의 어절 간 의존관계가 성립될 때 의존소와 지배소가 어떠한 관계를 가지는지 의존 관계명을 부착하는 모델을 제안한다. 국내에서 한국어 의존구문분석에 관한 연구가 활발히 진행되고 있지만 의존 관계만을 결과로 제시하고 의존 관계명을 제공하지 않는 경우가 많았다. 따라서 본 논문에서는 의존 경로(Dependency Path)와 음절의 의존 관계명 분포를 반영하는 음절 임베딩를 이용한 의존 관계명 부착 모델을 제안한다. 문장에서 나올 수 있는 최적의 입력 열인 의존 경로(Dependency Path)를 순차 레이블링에서 좋은 성능을 나타내고 있는 bidirectional LSTM-CRFs의 입력 값으로 사용하여 의존 관계명을 결정한다. 제안된 기법은 자질에 대한 많은 노력 없이 의존 경로에 따라 어절 및 음절 단어표상(word embedding)만을 사용하여 순차적으로 의존 관계명을 부착한다. 의존 경로를 사용하지 않고 전체 문장의 어절 순서를 바탕으로 자질을 추출하여 CRFs로 분석한 기존 모델보다 의존 경로를 사용했을 때 4.1%p의 성능향상을 얻었으며, 의존 관계명 분포를 반영하는 음절 임베딩을 사용한 bidirectional LSTM-CRFs는 의존 관계명 부착에 최고의 성능인 96.01%(5.21%p 개선)를 내었다.

• The Journal of Society for e-Business Studies
• /
• v.26 no.1
• /
• pp.29-41
• /
• 2021

Recently, there are increasing attempts to utilize deep learning methodology in the fashion industry. Accordingly, research dealing with various fashion-related problems have been proposed, and superior performances have been achieved. However, the studies for fashion style classification have not reflected the characteristics of the fashion style that one outfit can include multiple styles simultaneously. Therefore, we aim to solve the multi-label classification problem by utilizing the dependencies between the styles. A multi-label recognition model based on a graph convolution network is applied to detect and explore fashion styles' dependencies. Furthermore, we accelerate model training and improve the model's performance through transfer learning. The proposed model was verified by a dataset collected from social network services and outperformed baselines.

• Journal of the Korea Society of Computer and Information
• /
• v.17 no.7
• /
• pp.45-52
• /
• 2012

In this paper, an crowd density in elevator estimation method based on multi-class Adaboost classifier is proposed. The SOM (Self-Organizing Map) based conventional methods have shown insufficient performance in practical scenarios and have weakness for low reproducibility. The proposed method estimates the crowd density using multi-class Adaboost classifier with texture features, namely, GLDM(Grey-Level Dependency Matrix) or GGDM(Grey-Gradient Dependency Matrix). In order to classify into multi-label, weak classifier which have better performance is generated by modifying a weight update equation of general Adaboost algorithm. The crowd density is classified into four categories depending on the number of persons in the crowd, which can be 0 person, 1-2 people, 3-4 people, and 5 or more people. The experimental results under indoor environment show the proposed method improves detection rate by about 20% compared to that of the conventional method.