• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.20 no.6
• /
• pp.313-330
• /
• 2021

In an autonomous driving system equipped with a camera and lidar, depth completion techniques enable dense depth estimation. In particular, using self-supervised learning it is possible to train the depth completion network even without ground truth. In actual autonomous driving, such depth completion should have very short latency as it is the input of other algorithms. So, rather than complicate the network structure to increase the accuracy like previous studies, this paper focuses on network latency. We design a U-Net type network with RegNet encoders optimized for GPU computation. Instead, this paper presents several techniques that can increase accuracy during the process of self-supervised learning. The proposed techniques increase the robustness to unreliable lidar inputs. Also, they improve the depth quality for edge and sky regions based on the semantic information extracted in advance. Our experiments confirm that our model is very lightweight (2.42 ms at 1280x480) but resistant to noise and has qualities close to the latest studies.

• Journal of KIISE:Software and Applications
• /
• v.37 no.1
• /
• pp.60-73
• /
• 2010

This paper aims to describe KorLexClas 1.5 which provides us with a very large list of Korean numeral classifiers, and with the co-occurring noun categories that select each numeral classifier. Differently from KorLex of other POS, of which the structure depends largely on their reference model (Princeton WordNet), KorLexClas 1.0 and its extended version 1.5 adopt a direct building method. They demand a considerable time and expert knowledge to establish the hierarchies of numeral classifiers and the relationships between lexical items. For the efficiency of construction as well as the reliability of KorLexClas 1.5, we use following processes: (1) to use various language resources while their cross-checking for the selection of classifier candidates; (2) to extend the list of numeral classifiers by using a shallow parsing techniques; (3) to set up the hierarchies of the numeral classifiers based on the previous linguistic studies; and (4) to determine LUB(Least Upper Bound) of the numeral classifiers in KorLexNoun 1.5. The last process provides the open list of the co-occurring nouns for KorLexClas 1.5 with the extensibility. KorLexClas 1.5 is expected to be used in a variety of NLP applications, including MT.

• Korean Journal of Cognitive Science
• /
• v.19 no.2
• /
• pp.143-161
• /
• 2008

The goal of this paper is to extract the hyponymy relation between words in the corpus. Adopting the basic algorithm of Hearst (1992), I propose a method of pattern-based extraction of semantic relations from the corpus. To this end, I set up a list of hypernym-hyponym pairs from Sejong Electronic Dictionary. This list is supplemented with the superordinate-subordinate terms of CoroNet. Then, I extracted all the sentences from the corpus that include hypemym-hyponym pairs of the list. From these extracted sentences, I collected all the sentences that contain meaningful constructions that occur systematically in the corpus. As a result, we could obtain 21 generalized patterns. Using the PERL program, we collected sentences of each of the 21 patterns. 57% of the sentences are turned out to have hyponymy relation. The proposed method in this paper is simpler and more advanced than that in Cederberg and Widdows (2003), in that using a word net or an electronic dictionary is generally considered to be efficient for information retrieval. The patterns extracted by this method are helpful when we look fer appropriate documents during information retrieval, and they are used to expand the concept networks like ontologies or thesauruses. However, the word order of Korean is relatively free and it is difficult to capture various expressions of a fired pattern. In the future, we should investigate more semantic relations than hyponymy, so that we can extract various patterns from the corpus.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_1
• /
• pp.1329-1341
• /
• 2022

Destroying the marine environment and marine ecosystem and causing marine accidents, marine debris is generated every year, and among them, submerged marine debris is difficult to identify and collect because it is on the seabed. Therefore, deep-learning-based semantic segmentation was experimented on waste fish nets and waste ropes using underwater images to identify efficient collection and distribution. For segmentation, a high-resolution network (HRNet), a state-of-the-art deep learning technique, was used, and the performance of each optimizer was compared. In the segmentation result fish net, F1 score=(86.46%, 86.20%, 85.29%), IoU=(76.15%, 75.74%, 74.36%), For the rope F1 score=(80.49%, 80.48%, 77.86%), IoU=(67.35%, 67.33%, 63.75%) in the order of adaptive moment estimation (Adam), Momentum, and stochastic gradient descent (SGD). Adam's results were the highest in both fish net and rope. Through the research results, the evaluation of segmentation performance for each optimizer and the possibility of segmentation of marine debris in the latest deep learning technique were confirmed. Accordingly, it is judged that by applying the latest deep learning technique to the identification of submerged marine debris through underwater images, it will be helpful in estimating the distribution of marine sedimentation debris through more accurate and efficient identification than identification through the naked eye.

• Journal of the Korean Society for Library and Information Science
• /
• v.49 no.1
• /
• pp.45-67
• /
• 2015

The purpose of this study is to analyse the semantic relations of terms in STNet, a structured terminology dictionary based on author keywords of humanities, social sciences, and art and sport in the Korea Citation Index (KCI) and to describe the procedure for optimizing the relation types and specifying the name of relationships. The results indicate that four logical criteria, such as creating new names for relationships or limitation of typing the relationship by the appearance frequency of same type, consideration of direction of relationship, reflection to accept the existing name of relationships, are required for the optimization of the typing and naming the relationships. We applied these criteria to the relationships in the class "real person" of STNet and the result shows that 1,135 out of 1,743 uncertain relationships such as RT, RT_X or RT_Y are specified and clarified. This rate of optimization with ca. 65% represents the usefulness of the criteria applicable to the cases of database construction and retrieval.

• Annual Conference on Human and Language Technology
• /
• 2001.10d
• /
• pp.538-545
• /
• 2001

본 연구에서는 오픈 도메인에서 동작할 수 있는 질의 응답 시스템(Open-domain Question Answer ing System)을 구현하고 영어권 TREC에 참가한 결과를 기술하였다. 정답 유형을 18개의 상위 노드를 갖는 계층구조로 분류하였고, 질문 처리에서는 LSP(Lexico-Semantic Pattern)으로 표현된 문법을 사용하여 질문의 정답 유형을 결정하고, lemma 형태와 WordNet 의미, stem 형태의 3가지 유형의 키워드로 구성된 질의를 생성한다. 이 질의를 바탕으로, 패시지 선택에서는 문서검색 엔진에 의해 검색된 문서들을 문장단위로 나눠 정수를 계산하고, 어휘체인(Lexical Chain)을 고려하여 인접한 문장을 결합하여 패시지를 구성하고 순위를 결정한다. 상위 랭크의 패시지를 대상으로, 정답 처리에서는 질문의 정답 유형에 따라 품사와 어휘, 의미 정보로 기술된 LSP 매칭과 AAO (Abbreviation-Appositive-Definition) 처리를 통해 정답을 추출하고 정수를 계산하여 순위를 결정한다. 구현된 시스템의 성능을 평가하기 위해 TREC10 QA Track의 main task의 질문들 중, 200개의 질문에 대해 TRIC 방식으로 자체 평가를 한 결과, MRR(Mean Reciprocal Rank)은 0.341로 TREC9의 상위 시스템들과 견줄 만한 성능을 보였다.

• KIISE Transactions on Computing Practices
• /
• v.22 no.2
• /
• pp.107-112
• /
• 2016

The Sejong Electronic(machine-readable) Dictionary, developed by the 21st century Sejong Plan, contains systematically organized information on Korean words. It helps to solve problems encountered in the electronic formatting of the still-commonly-used hard-copy dictionary. The Sejong Electronic Dictionary, however has a limitation relate to sentence structure and selection-restricted nouns. This paper discuses the limitations of word-sense disambiguation(WSD) that uses subcategorization information suggested by the Sejong Electronic Dictionary and generalized selection-restricted nouns from the Korean Lexico-semantic network. An alternative method that utilized semi-supervised learning, the chi-square test and some other means to make WSD decisions is presented herein.

• Journal of KIISE:Computer Systems and Theory
• /
• v.33 no.9
• /
• pp.563-573
• /
• 2006

Cyber-infrastructures for scientific and engineering applications require integrating heterogeneous legacy data in different formats and from various domains. Such data integration raises challenging issues: (1) Support for multiple independently-managed schemas, (2) Ease of schema evolution, and (3) Simple schema mappings. In order to address these issues, we propose a novel approach to semantic integration of scientific data which uses XML schemas and RDF-based schema mappings. In this approach, XML schema al-lows scientists to manage data models intuitively and to use commodity XML DBMS tools. A simple RDF-based ontological representation scheme is used for only structural relations among independently-managed XML schemas from different institutes or domains We present the design and implementation of a prototype system developed for the national cyber-environments for civil engi-neering research activities in Korea (similar to the NEES project in USA) which is called KOCEDgrid (http://www.koced.net).

• KIISE Transactions on Computing Practices
• /
• v.21 no.7
• /
• pp.500-505
• /
• 2015

The Sejong Electronic(machine readable) Dictionary, which was developed by the 21 century Sejong Plan, contains a systematic of immanence information of Korean words. It helps in solving the problem of electronical presentation of a general text dictionary commonly used. Word sense disambiguation problems can also be solved using the specific information available in the Sejong Electronic Dictionary. However, the Sejong Electronic Dictionary has a limitation of suggesting structure of sentences and selection-restricted nouns. In this paper, we discuss limitations of word sense disambiguation by using subcategorization information as suggested by the Sejong Electronic Dictionary and generalize selection-restricted noun of argument using Korean Lexico-semantic network.

• Fashion & Textile Research Journal
• /
• v.21 no.3
• /
• pp.277-287
• /
• 2019

This study analyzes the online issues of design innovation by a fashion creative director. The study selected fashion house Gucci as the main subject and analyzed social media posts. As for study methods, a social matrix program Textom 2.0 collected 13,014 nouns and adjectives using 'Gucci Collection' as a search keyword from Naver Blogs from March to August 2014 and from March to August 2016. Design related issues were derived through semantic network analysis using Ucinet6 and the NetDraw program. The results of the keyword frequency analysis showed that social media user interest for the Gucci collection increased based on the rapid increase in the number of posts from 1,064 to 2,126 after changing the fashion creative director. The results of visualization of semantic network analysis and content analysis also showed that the main issues related to the Gucci collection design changed after the replacement of the fashion creative director. The study found that issues formed around the product information worn by celebrities for promotion purposes during the 2014 period; however, during the 2016 period, issues were formed around 'vintage' and 'retro' runway concepts with design styles related to Alessandro Michele, the new creative director.