• Journal of the Korea Convergence Society
• /
• v.13 no.1
• /
• pp.13-21
• /
• 2022

Recent deep learning-based research shows excellent performance in most natural language processing (NLP) fields with pre-trained language models. In particular, the auto-encoder-based language model proves its excellent performance and usefulness in various fields of Korean language understanding. However, the decoder-based Korean generative model even suffers from generating simple sentences. Also, there is few detailed research and data for the field of conversation where generative models are most commonly utilized. Therefore, this paper constructs multi-turn dialogue data for a Korean generative model. In addition, we compare and analyze the performance by improving the dialogue ability of the generative model through transfer learning. In addition, we propose a method of supplementing the insufficient dialogue generation ability of the model by extracting recommended response candidates from external knowledge information through a retrival model.

• Korean Journal of Heritage: History & Science
• /
• v.54 no.3
• /
• pp.4-23
• /
• 2021

The purpose of this study was to classify the types of tweezers used across time and to examine the social culture of periods through the use of tweezers. Chapter II summarized the terms for tweezers and looked at the use of tweezers in two cases related to the social culture and politics of a period. The first is the use of tweezers for beauty purposes. Men used tweezers as a kind of self-management to maintain their social status and power, and thus they helped develop a wide range of tweezers cultures. People with gray hair are usually old. The perception that we should step down from politics when we are aged has become strong. Therefore, politicians pulled out gray hairs with tweezers to maintain only black hair and show youth, which indicates that their social influence is still strong. The second is the use of tweezers for first aid. Chapter III classified tweezers by type during the Goryeo and Joseon periods. The tweezers are largely divided into basic and composite types. The basic type of tweezers is a form that has been continued since the Three Kingdoms Period and consists of tongs and handles, and a fixed type of tweezers with a fixed ring appeared. Composite tweezers are made by grafting earpick or multipurpose knives. Composite tweezers are all-in-one tweezers with an earpick and an all-in-one knife. Tweezers are usually all-in-one. Among the composite tweezers, rivet separation and ring separation are characterized by separation, unlike the all-in-one tweezers. The method of connecting is divided into rivets and connection rings. The all-in-one tweezers appear only during the Goryeo Dynasty and are characterized by the lifestyle that provides a glimpse of the tastes of contemporary users. The manufacturing takes shape after making a thin metal plate. Decorative techniques are carved on soft metal tweezers, such as silver and bronze with a line, point, and a pressed angle. These tweezers are presumed to have been used by the royal family or aristocrats. However, most tweezers are made of strong bronze or iron. Therefore, the majority of simple X-shaped patterns are sampled or without patterns. The biggest reason why there are such diverse types of tweezers is that the culture of tweezers was widespread regardless of the times. In addition, the basic type of tweezers has been used since the Three Kingdoms Period and has been modified and used together as necessary because the shape of tweezers is a practical daily tool. Study of metal crafts have been limited to royal objects and Buddhist crafts. We hope that research on everyday tools such as tweezers will continue to serve as an opportunity to examine the social and cultural aspects of the times in various ways.

• Koreanishche Zeitschrift fur Deutsche Sprachwissenschaft
• /
• v.6
• /
• pp.241-261
• /
• 2002

In der Grammatik $f\"{u}r$ die Lernenden werden die deutschen Tempora nach dem lateinischen Vorbild dargestellt: $Pr\"{a}sens,\;Pr\"{a}teritum$, Perfekt, Plusquamperfekt, Futur I und Futur II. Und diese sechs Tempora werden in die drei Zeitstufen kIassifiziert: Gegenwart, Vergangenheit und Zukunft. Aber die Zeitformen $k\"{o}nnen$ nicht in direkter und geradliniger Weise auf bestimmte objektive Zeiten beziehen. Das $Verh\"{a}ltnis$ zwischen Zeitinhalt und Zeitform ist verwickelter und komplexer. Man denkt nach, wie viele Tempora es in der deutschen Sprache gibt und warum die Tempora Funktionen, die nicht nur Zeitinhalt, sondern auch andere Inhalte(z.B. modale) $ausdr\"{u}ken$, haben. Urn die Probleme zu $erl\"{a}utern$, sollen wir die deutschen Tempora von der lateinischen Tempuskategorie abtrennen und ein geeignetes Tempussystem $f\"{u}r$ die deutsche Sprache suchen. Dazu $m\"{u}ssen$ wir beobachten, Wie die deutschen Tempora konstruiert wurden. Im Lateinischen handelt es sich bei allen sechs Tempora um Verbformen, um bestimmte ($n\"{a}mlich$ die indikativischen) Formen des finiten Verbs. Im Deutschen aber finden wir unter den Tempora nur zwei richtige Verbformen $(Pr\"{a}sens\;und\;Pr\"{a}teriturn),\;die\;\"{u}brigen$ Tempora sind verbale Komplexe, drei zweiwortige (Perfekt, Plusquamperfekt, Futur I) und ein dreiwortiger (Futur II). Und das deutsche Tempussystem besteht aus den zwei Polen: Vergangenes und Nicht-Vergangenes oder $Temporalit\"{a}t\;und\;Atemporalit\"{a}t$. Die Tempusform $Pr\"{a}teritum\;geh\"{o}rt\;zu\;Temporalit\"{a}t$ und die Tempusform $Pr\"{a}sens\;zu\;Atemporalit\"{a}t$. Das $Pr\"{a}sens\;und\;das\;Pr\"{a}teritum$ sind die Grundtempora im Deutschen und die $\"{u}brigen$ Tempora sind die durch den Aspekt erweiterten und die zusammengesetzten Tempora. Sie haben deswegen die inneren Merkmale Aspekt und Modale $au{\ss}er$ Zeitbezug. Wir $k\"{o}nnen$ diese inneren Merkmale im Tempusgebrauch feststellen. Als Funktionen der Tempora der deutschen Sprache im Tempusgebrauch sind zu registrieren: Bezeichnung zeitlicher Gegebenheit, modale Implikationen, Aspektangaben und semantische Akzentuierungen. Und noch eins: Die Funktion des neutralen Referieren des Inhaltes durch das atemporale $Pr\"{a}sens$. Das $Pr\"{a}sens$ wird durch diese Referentfunktion zu dem Haupttempus auch in der geschriebenen Sprache. Trotzdem nennt man das $Pr\"{a}teritum$ in der geschriebenen Sprache das Haupttempus. Ist das $Pr\"{a}teritum$ dann in der Tat in der geschriebenen Sprache das Haupttempus? Genau genommen ist das $Pr\"{a}sens$ offenbar ein Haupttempus. Es wird nicht nur in der gesprochenen Sprache, sowohl auch in den geschriebenen nichtbelletristischen Texten $haupts\"{a}chlich$ benutzt. Aber nur die belletristischen Texte werden durch das $Pr\"{a}teritum$ beherrscht. Der Gebrauch des $Pr\"{a}teritums$ als Haupttempus $beschr\"{a}nkt$ sich nur auf die Belletristik. Wir glauben, die Untersuchung $\"{u}ber$ den Tempusgebrauch und die Tempusfunktion in den verschiedenen Textsorten sei besonders dem Nichtmuttersprachler wichtig, weil die Textsorten beim Tempusgebrauch eine Wahl des Tempus beeinflussen.

• 한국방재학회:학술대회논문집
• /
• 2011.02a
• /
• pp.153-153
• /
• 2011

현대 건축은 기능적이고 합리적이었지만 획일적이었던 박스형 건축에서 탈피해 형태와 공간에 있어서 다양한 변화를 시도하고 있다. 특별한 건축물의 실현을 위해 각 나라의 기술력은 급속한 발전을 이루었고, 보다 더 독특한 건축물에 대한 관심은 비정형 건축물에 대한 관심의 증대로 이어지고 있다. 이러한 비정형 건축물에 적합한 구조로써 프리폼(Free-Form) 구조가 있다. 프리폼 구조로 입체골조(Double Layered Structure)를 많이 사용하였으나, 최근 유리로 되어 투명하고 기하학적 모양의 건축물을 추구함에 따라 평면골조(Single Layered Structure)가 증가하고 있는 추세이다. 평면골조는 축력 지배형과 모멘트 지배형으로 분류할 수 있고 프리폼 구조의 구성 요소 중 가장 취약하고 중요한 부분은 노드이다. 본 연구에서는 프리폼 구조 중 가장 큰 관심이 고조되고 있는 평면골조 모멘트 지배형의 노드에 대한 국내외 기술 분석을 통해 향후 연구 방향성을 제시하고자 한다. 입체골조는 하나의 노드에 여러개의 부재가 3차원으로 결합되어야 하기 때문에 다른 골조 시스템에 비해 노드부가 복잡하지만, 건축물의 외관을 유리로 하여 투명하게 하고 비틀리고 구부러진 구조물에 대한 건축적 요구가 많아짐에 따라 평면골조의 인기가 높아지고 있다. 이러한 시대의 흐름에 발맞추어 건물의 구조적, 기하학적 요구를 충족시키기 위해 다양한 노드 시스템이 개발 중이며, 가해지는 하중의 특성에 따라 축력 지배형과 모멘트 지배형으로 구분하여 노드의 양상을 분류할 수 있다. 축력 지배형의 대표적인 시스템은 다이아그리드(Diagrid)이다. 축력 지배형 프리폼 구조의 노드는 전체 구조물의 하중을 축력으로 받아 모두 전달해야 하기 때문에 크기가 크고 가새가 2~4개층에 걸쳐서 설치되기 때문에 중량이다. 모멘트 지배형 노드를 갖는 프리폼 구조의 형태는 대부분 지붕 구조로써 지붕 자체의 하중만을 견디도록 설계된다. 따라서 노드부와 노드에 붙는 부재들이 가볍기 때문에 사람이 들 수 있고 노드의 크기가 작아 시공성이 좋으며 대량 생산이 가능하다는 장점이 있다. 노드의 형태는 힘의 흐름과 쓰임에 따라 다양하다. 평면골조 모멘트 지배형의 노드는 접합방식에 따라 Splice node connection과 End-Face node connection 두 가지로 분류할 수 있다. Splice node connection은 각 부재의 종축으로 노드와 구조부재 사이에 이음재를 두어 연결하고, 연결 형태에 따라 전단력을 전달할 수 있는 1~2개의 접촉면이 생긴다. 전단응력을 받는 볼트로 이음재를 이어 조립하거나 용접으로 접합할 수 있다. 대표 노드로, SBP-1, SBP-2와 POLO-1 등이 있다. End-Face node connection은 각 연결된 부재의 단부와 노드 사이의 연결면은 종축방향의 수직이고, 인장응력을 받는 볼트를 사용하거나 용접에 의해 접합할 수 있다. 대표 노드로 SBP-4, WABI-1, MERO-1(Cylinder), MERO-2(Block), MERO-4(Double Dish) 등이 있다. 본 기술 현황 분석을 통해 현재 개발된 노드를 분류하고 가장 관심이 높은 Single Layer 모멘트 지배형 노드를 비교, 분석하였다. 최근 건물의 경향을 반영한 프리폼 구조를 실현하기 위해서 필수적인 노드의 개발은 국외에서 활발히 연구되고 있지만 그 기술이 개방되어 있지 않다. 국내에서는 동대문 디자인 플라자에 새로운 노드를 적용하고 고려대학교에서 모멘트 지배형 노드를 개발하는 등 발전 가능성을 보이고 있지만 국외 사례들에 비하면 아직 초기 단계라 할 수 있다. 따라서 현장 용접을 지양하고 공장 제작하여 현장에서 조립하며, 프로젝트 별로 상이한 노드를 사용하는 것이 아닌 다양한 요구를 효과적으로 수용하는 구조 효율성을 향상시킨 노드 상세의 개발이 이루어져야 할 것이다.

• Korean journal of communication and information
• /
• v.43
• /
• pp.314-358
• /
• 2008

This study examined how the meaning of news stories covering the split of national opinion was constructed in the media to represent social conflicts. To clarify the function of the term 'split of national opinion' as an ultimate term, this study examined the meaning of the term in the context of both text and society. Ten newspapers were included in the content analysis. The frequency of words used for the purpose of metaphor and equivalent in describing the split of national opinion was calculated to determine their meaning in the textual context. The frequency of incidents and subjects involved in allegedly causing the split of national opinion was calculated to determine their meaning in the social context. The results of this study are summarized as follows: First, the term 'split of national opinion' was coined by the newspapers as a metaphor of disease, disaster, and cost. The attitudes or the ways in which the split of national opinion was dealt with were generally negative and passive. Second, the term 'split of national opinion' was dealt with an equivalent status of such terms as national policy, national loss, societal problems, and ideology. Third, each newspaper reported that the split of national opinion had been caused by certain subjects, which indicates that each newspaper had its own position of viewing who was the key player in splitting the national opinion. The implication was also discussed that the use of the ultimate term would incur the unbalance of power between participants and the existing players, which would make individuals or groups who were involved in the social actions excluded and make the newspapers exercise the rhetorical power as news media.

• The Journal of the Acoustical Society of Korea
• /
• v.32 no.6
• /
• pp.548-555
• /
• 2013

The localization of sources has a numerous number of applications. To estimate the position of sources, the relative delay between two or more received signals for the direct signal must be determined. Although the generalized cross-correlation method is the most popular technique, an approach based on eigenvalue decomposition (EVD) is also popular one, which utilizes an eigenvector of the minimum eigenvalue. The performance of the eigenvalue decomposition (EVD) based method degrades in the low SNR and the correlated environments, because it is difficult to select a single eigenvector for the minimum eigenvalue. In this paper, we propose a new adaptive algorithm based on Canonical Correlation Analysis (CCA) in order to extend the operation range to the lower SNR and the correlation environments. The proposed algorithm uses the eigenvector corresponding to the maximum eigenvalue in the generalized eigenvalue decomposition (GEVD). The estimated eigenvector contains all the information that we need for time delay estimation. We have performed simulations with uncorrelated and correlated noise for several SNRs, showing that the CCA based algorithm can estimate the time delays more accurately than the adaptive EVD algorithm.

• Journal of Conservation Science
• /
• v.29 no.2
• /
• pp.171-185
• /
• 2013

The purpose of this study were to interpret making characteristics and clay sources for Proto-Three Kingdoms period potteries which were excavated from Bakjimeure site in Tangjeong, Asan, Korea, on the basis of shape types. The pottery samples of the studied included deep bowl pottery, cylinder shaped pottery, duck shaped pottery and rounded based jar with short neck. And the last jar was dominant among them. This is supported in nearby Yongduri Jinter site, Galmaeri site and Baekseokdong site. They vary in terms of color of clay, absorption ratio and mineral classification. Based on the fact that their absorption ratio range was 4.50 to 25.19% and firing temperature range was 850 to $1,150^{\circ}C$, they were thought to be widely used. The equivalence of cylinder shaped pottery, duck shaped pottery and deep bowl pottery was high in terms of material characteristics and their firing temperature range was 850 to $950^{\circ}C$. However, the deep bowl pottery, compared to other pottery shapes, contained great amount of coarse grained quartz and feldspar, and was different from others in terms of function. For soil sample collected from relative plains (area 3) in the southernmost part of the site, their mineralogical and geochemical characteristics were the most similar to those of the potteries excavated from the Bakjimeure site and even most of the potteries collected from different sites.

• Smart Media Journal
• /
• v.7 no.1
• /
• pp.16-23
• /
• 2018

In this paper, we used the Deep Neural Network (DNN) to predict the number of daily spectators of Gwangju - KIA Champions Field in order to provide marketing data for the team and related businesses and for managing the inventories of the facilities in the stadium. In this study, the DNN model, which is based on an artificial neural network (ANN), was used, and four kinds of DNN model were designed along with dropout and batch normalization model to prevent overfitting. Each of four models consists of 10 DNNs, and we added extra models with ensemble model. Each model was evaluated by Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The learning data from the model randomly selected 80% of the collected data from 2008 to 2017, and the other 20% were used as test data. With the result of 100 data selection, model configuration, and learning and prediction, we concluded that the predictive power of the DNN model with ensemble model is the best, and RMSE and MAPE are 15.17% and 14.34% higher, correspondingly, than the prediction value of the multiple linear regression model.

• Korean Journal of Dental Materials
• /
• v.45 no.4
• /
• pp.301-310
• /
• 2018

The purpose of this study is to evaluate the compatibility of universal adhesives with two dual-cured resin cements. Eighty human molars were divided into eight groups. Tooth was embedded in self-curing acrylic resin and sectioned horizontally to exposure dentine surface. After polishing with 600-grit SiC paper, adhesives were applied. All-Bond Universal (Bisco), G-premio bond (GC), Scotch-bond universal (3M ESPE), Scotch-bond Multipurpose (3M ESPE) were used in this study. Calibra (Dentsply) as a conventional dual-curing resin cement and RelyX Ultimate (3M ESPE) as an amine free resin cement were used. The adhesives and the cements were applied according to the manufacturer's instructions. Final specimens were cylinder (diameter 2mm, height 3mm) shape. After storing in distilled water at $37^{\circ}C$ for 7 days the shear bond strength (SBS) test was performed. There was no significant difference in shear bond strength between the adhesives when RelyX Ultimate was used (p>0.05). However, when Calibra used with Scotch-bond Multipurpose and All-Bond Universal were used, statistically higher SBS was observed, as compared to the groups which Calibra cements with G-premio bond and Scotch-bond universal adhesive (p<0.05) were used. Within the limitations of this study, RelyX Ultimate resin cement was compatible with universal adhesives of various pH. All-Bond Universal adhesive was compatible with a resin cement containing the tertiary amine.

• Korea journal of population studies
• /
• v.35 no.2
• /
• pp.55-72
• /
• 2012

Few studies have done on the intergenerational relations of married women and their own parents in Japan. This study approaches the topic by examining coresidence and contact frequency between generations. The study expects the likelihood of living together (including living next door) and the extent of contact would differ by the characteristics of woman, her husband, children, her brothers and sisters, her own parents and parents-in-law. From the 2003 Survey for National Family Research in Japan, selected are 853 currently married women in their 30s and 40s whose parent and parents-in-law are alive. The analysis shows that the likelihood of living together with parents decreases as the number of brothers and sisters increases. In particular, the presence of brother substantially decreases the likelihood. Having father only alive (vs. having both parents alive) also increases the likelihood. The frequency of contact with parents is conditioned by the coresidence with parents-in-law. It also differs by the level of education and its gap between spouses. Subjective evaluation of husband's attitude toward her parents is important. As in the case of living together, the number of brothers and sisters and the survival status of parents are significant in explaining the frequency of contact with her parents. The results indicate that number of brothers and sisters as well as widowhood of parents serves as its demographic condition. The findings that the frequency of contact with parent are affected by coresidene with parents-in-law, education gap between spouses and husband's attitude toward her parents suggest that the relationship of married women with her own parents are conditioned by her husband and his parents.