• 한국빅데이터학회지
• /
• 제7권2호
• /
• pp.11-29
• /
• 2022

사전학습 언어모델은 자연어 처리 작업에서 가장 중요하고 많이 활용되는 도구로, 대량의 말뭉치를 대상으로 사전학습이 되어있어 적은 수의 데이터를 이용한 미세조정학습으로도 높은 성능을 기대할 수 있으며, 사전학습된 토크나이저과 딥러닝 모형 등 구현에 필요한 요소들이 함께 배포되기 때문에 자연어 처리 작업에 소요되는 비용과 기간을 크게 단축시켰다. 트랜스포머 변형 모형은 이와 같은 장점을 제공하는 사전학습 언어모델 중에서 최근 가장 많이 사용되고 있는 모형으로, 번역을 비롯하여 문서 요약, 챗봇과 같은 질의 응답, 자연스러운 문장의 생성 및 문서의 분류 등 다양한 자연어 처리 작업에 활용되고 있으며 컴퓨터 비전 분야와 오디오 관련 분야 등 다른 분야에서도 활발하게 활용되고 있다. 본 논문은 연구자들이 보다 쉽게 사전학습 언어모델에 대해 이해하고 자연어 처리 작업에 활용할 수 있도록 하기 위해, 언어모델과 사전학습 언어모델의 정의로부터 시작하여 사전학습 언어모델의 발전과정과 다양한 트랜스포머 변형 모형에 대해 조사하고 정리하였다.

• Asian-Australasian Journal of Animal Sciences
• /
• 제22권8호
• /
• pp.1069-1077
• /
• 2009

Castration of male pig produces significant negative effects on skeletal muscle development. The androgen receptor (AR), two splice variants of insulin-like growth factor-I (IGF-I Ea and MGF) and the myostatin gene may play important roles in this process. In the present study, the expression of AR, IGF-I Ea, MGF and myostatin genes in three skeletal muscles, the brachialis, longissimus and semitendinosus, were studied using real-time quantitative RT-PCR. Our experimental design used 14 pairs of male Landrace sire${\times}$Yorkshire dam piglets. The two piglets in each pair were full sibs, one of which was castrated at 21 d of age; the other remained intact. The study group was divided into subgroups of equal size. Animals in the first subgroup were slaughtered at 147 d and those of the second at 210 d of age. Carcass weight and lean meat yield were similar between boars and barrows at 147 d of age (p>0.05), whereas barrows had lower carcass weight and less lean meat yield at 210 d of age (p<0.05). Castration caused down-regulation of AR gene expression at both 147 and 210 d of age (p<0.05). The two splice variants of the IGF-I gene from porcine skeletal muscle were cloned using RT-PCR, and it was found that MGF differs from IGF-I Ea in having a 52-base insert in the last coding exon of the mRNA. Both splice variants were down-regulated by castration only at 210 d of age (p<0.05). No differences in expression of the myostatin gene were observed between boars and barrows at either 147 or 210 d of age (p>0.05). These results suggest that the downregulation of AR, IGF-I Ea and MGF gene expression following castration helps to explain the negative effect of castration on skeletal muscle development.

• 한국수학교육학회지시리즈B:순수및응용수학
• /
• 제27권4호
• /
• pp.187-205
• /
• 2020

In this article, we prove coincidence point theorems for comparable 𝜓-contractive mappings in ordered non-Archimedean fuzzy metric spaces utilizing the recently established concept of 𝓣-comparability and relatively weaker order theoretic variants. With a view to show the usefulness and applicability of this work, we solve the system of ordered Fredholm integral equations as an application. In the process, this presentation generalize and improve some prominent recent results obtained in Mihet [Fuzzy Sets Syst., 159 (6), 739-744, (2008)], Altun and Mihet [ Fixed Point Theory Appl. 2010, 782680, (2010)], Alam and Imdad [Fixed Point Theory, 18(2), 415-432, (2017)] and several others in the settings of partially ordered non-Archimedean fuzzy metric spaces.

• 한국수학교육학회지시리즈D:수학교육연구
• /
• 제8권4호
• /
• pp.261-277
• /
• 2004

Julia sets, their variants and generalizations have been studied extensively by using the Picard iterations. The purpose of this paper is to introduce Mann iterative procedure in the study of Julia sets. Escape criterions with respect to this process are obtained for polynomials in the complex plane. New escape criterions are significantly much superior to their corresponding cousins. Further, new algorithms are devised to compute filled Julia sets. Some beautiful and exciting figures of new filled Julia sets are included to show the power and fascination of our new venture.

• 음성과학
• /
• 제15권1호
• /
• pp.53-62
• /
• 2008

The purpose of this study was to examine the English l sound in an English short story produced by 16 Korean students in order to determine various allophones of the sound using acoustic visual displays and perceptual judgments. The subjects read the story in a quiet office at normal speed. Each word included the lateral sound in onset or coda positions and before a vowel of the following word. Results showed as follows: Firstly, there was a durational difference between the two major groups. Also the majority of the subjects produced the clear l regardless of the contexts. Some students produced the sound as the Korean flap or the English glide [r]. A few missing cases were also seen. The dark l was mostly produced by the subjects of English majors in coda position with a few cases before a vowel in a phrase. Visual displays using the computer analysis were very helpful in distinguishing lateral variants but sometimes perceptual process would be necessary to judge them in fast and weak production of the target word. Further studies would be desirable to test the discrepancies between the acoustical and perceptual decisions.

• 말소리와 음성과학
• /
• 제8권4호
• /
• pp.9-14
• /
• 2016

This study investigated the patterns of coda deletion in spontaneous Seoul Korean speech. More specifically, the current study focused on three factors in promoting coda deletion, namely, word position, consonant type, and morpheme type. The results revealed that, first, coda deletion frequently occurred when affixes were attached to the ends of words, rather than in affixes in word-internal positions or in roots. Second, alveolar consonants [n] and [l] in the coda positions of high-frequency affixes [nɨn] and [lɨl] were most likely to be deleted. Additionally, regarding affix reduction in the word-final position, all subjects seemed to depend on this articulatory strategy to a similar degree. In sum, the current study found that affixes without primary semantic content in spontaneous speech tend to undergo the process of reduction, favoring the occurrence of specific pronunciation variants.

• Journal of the Korean Data and Information Science Society
• /
• 제15권4호
• /
• pp.783-791
• /
• 2004

In this paper we review recent developments in nonlinear time series modeling on both conditional mean and conditional variance. Traditional linear model in conditional mean is referred to as ARMA(autoregressive moving average) process investigated by Box and Jenkins(1976). Nonlinear mean models such as threshold, exponential and random coefficient models are reviewed and their characteristics are explained. In terms of conditional variances, ARCH(autoregressive conditional heteroscedasticity) class is considered as typical linear models. As nonlinear variants of ARCH, diverse nonlinear models appearing in recent literature including threshold ARCH, beta-ARCH and Box-Cox ARCH models are remarked. Also, a class of unified nonlinear models are considered and parameter estimation for that class is briefly discussed.

• 한국도서관정보학회지
• /
• 제32권1호
• /
• pp.249-260
• /
• 2001

This paper presents the application of best match search algorithm in the client/server system for natural language access to Web-based database. For this purpose, the procedures to process Korean word variants as well as to execute probabilistic weighting scheme have been implemented in the client/server system. The experimental runs have been done using a Korean test set which included documents, queries and relevance judgements. The experimental results demonstrate that best match retrieval with relevance information is better than the retrieval without it.

• 소성∙가공
• /
• 제8권6호
• /
• pp.604-611
• /
• 1999

This paper is concerned with the family of parts that generally feature a central hub with radial protrusions. As opposed to conventional forward and backward extrusion, in which the material flows in a direction parallel to that of the punch or die motion, the material flows perpendicular to the punch motion in radial extrusion. Three variants of radial extrusion of a collar or flange are investigated. Case I involves forcing a cylindrical billet against a flat die, Case II involves deformation against a stationary punch recessed in the lower die, and Case III involves both the upper and lower punches moving together toward the center of the billet. Extensive simulational work is performed with each case to see the process conditions in terms of forging load, balanced and symmetrical flow in the flange. Also, the effect of the gap size and die corner radii to the material flow are investigated. In this study, the forming characteristics of radial extrusion will be considered by comparing the forces, shapes etc. The design factors during radial extrusion are investigated by the rigid-plastic FEM simulation.

• 한국발생생물학회:학술대회논문집
• /
• 한국발생생물학회 2001년도 후기 제12차 학술대회 논문집
• /
• pp.23-25
• /
• 2001

1. Leptin itself was not expressed in mouse uterine tissues. 2. Leptin receptors were not expressed in nonpregnant and little expressed in 3.5 day of pregnant uterine tissues. However, there was a signal in 4.5 and 5.5 day of tissues. 3. The expression level of leptin receptor variants in the implantation sites at around the time of initial embryo attachment (day 4.5 of pregnancy) and during the actual implantation period (day 5.5 of pregnancy) was much lower than that in the interimplantation 4. Finding of the differential expression of leptin receptors in implantation sites compared to interimplantation sites suggests that leptin - leptin receptor system may be one of the delicate regulators in the molecular mechanism of the implantation process.