• Proceedings of the Korean Information Science Society Conference
• /
• 2001.10b
• /
• pp.181-183
• /
• 2001

정렬(alignment)은 병렬 코퍼스에서 원문서의 문단, 문장, 혹은 단어와 같은 단위 요소에 대해, 대역문서에서의 상응하는 단위 요소를 찾는 일로, 코퍼스 기반 기계번역 방식에서 매우 중요한 과정이다. 동일 어족간의 원문과 대역문에서는 어순이나 단위 요소들이 거의 일치하여 정렬에 큰 어려움이 없으나, 한국어와 영어와 같이 어족이 다른 언어간의 정렬은 언어의 단위 요소의 상이성과 어순의 차이 등으로 인해 않은 어려움이 존재한다. 본 논문은 어족이 다른 언어 사이의 정렬을 위해 상대 구문 고립성(Relative Syntactic Isolativity)이라는 개념을 적용하여 언어 단위의 상이성을 극복할 수 있는 단위 구를 제안하고 이들을 추출하는 방법에 대해 보인다.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.45 no.6
• /
• pp.463-472
• /
• 2017

Inertial Navigation System Alignment is the process to determine direction cosine matrix which is the transformation matrix between the INS body frame and navigation frame. INS initial position value is necessary to INS attitude calculation, so that user should wait until he get such value to start the INS alignment. To remove the waiting time, we propose an alignment algorithm that immediately starts after the INS power on by using pseudo initial position input and then is completed with attitude error compensation by entering true position later. We analyse effect of INS sensor error on attitude in process of time and verify the performance and usefulness of the close-loop alignment algorithm which corrects attitude error from the change of initial position.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.46 no.3
• /
• pp.1-9
• /
• 2009

Using the projected 2-dimensional tilted images obtained from the transmission electron microscopy, we can reconstruct the 3-dimensional structures of objects, such as cells. As a preprocessing procedure, the tilted images should be registered and compensated in terms of the spatial position and the intensity difference, respectively. In this paper, we employ the fiducial marker-based approach to perform a registration, and introduce a simple intensity compensation scheme. Based on the transmissivity image formation model, we propose an algorithm that can compensate the components of the illumination and transmissivity of each image according to the tilted angle. Numerical analysis using real images obtained from the transmission electron microscopy are shown in this paper to show the performance of the proposed algorithm.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2001.10a
• /
• pp.575-578
• /
• 2001

원격 조작이나 해저 수중 탐사에서 실제와 같이 거리감을 인식하며 제어하기 위하여, 3차원 입체 영상 카메라 장치를 사용하고 있다. 기본적인 구성 형태인 Stereoscopic Camera 시스템은 기구적으로 주시각 제어와 초점 제어가 이루어 지고 있으며, 획득영상에 대하여 영상 distortion 보정, 압축 처리 등의 영상 신호처리가 행하여진다. 양안 카메라의 수평 위치를 일치시켜 수평적으로 동일 위치의 pixel들이 정확한 epipolar line을 형성할 경우에, 주시각 제어가 용이하고 보정 및 영상처리 등의 연산량이 대폭 감소된다. 이와 같은 calibration 과정을, 기존의 시스템에서는, 주로 영상 획득 포기에 패턴을 사용하여 실시하거나, 물리적 수평 장치와 sensor 등의 보조 장치를 이용하여 calibration을 행한다. 그러나, 기계적으로 정밀하게 정렬을 한다고 하여도 두 카메라의 광축 및 CCD조립상 상이점과 특성의 불일치로 인하여 실제 획득된 영상에서는 변이와 회전이 포함된 영상을 얻게된다. 본 논문에서는 Stereoscopic Camera의 위와같은 정렬 오류의 문제점을 분석한 후, 제안 방식으로서 두 카메라의 획득되는 영상을 직접 영상 처리하여 수직 방향 및 회전 오류를 최소화 시켜 정렬하는 새로운 방법을 제시하며, 실험적으로 제안 방식의 효율성을 보인다.

• The Korean Journal of Pesticide Science
• /
• v.10 no.3
• /
• pp.157-164
• /
• 2006

Three dimensional quantitative structure-activity relationships (3D-QSARs) for the fungicidal activities against Rhizoctonia solani (RS) and Phytophthora capsici (PC) by N-phenyl substituents(X) of N-phenylthionocarbamate derivatives were studied quantitatively using comparative molecular field analysis (CoMFA) methodology based on different alignment approaches. Statistical quality of CoMFA models with field fit alignment were slightly higher than that of atom based fit alignment. The optimized CoMFA models (RS: RF2 & PC: PF2) were derived from field fit alignment and combination of CoMFA fields. And the statistical results of the two models showed the best predictability of the fungicidal activities based on the cross-validated value $q^2$ ($r^2_{cv.}$ =RS: 0.557 & PC: 0.676) and non-cross-validated value ($r^2_{ncv.}$ =RS: 0.954 & PC: 0.968), respectively. The selective fungicidal activities between two fungi were dependence upon the electrostatic field of substrate molecule. Therefore, the fungicidal activities from CoMFA contour maps showed that the fungicidal activity will be able to increased according to the modification of X-substituents on the substrate molecules.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2012.04a
• /
• pp.123-126
• /
• 2012

고도 컴퓨팅용 응용 프로그램 작성시 "에너지"를 감안하여 가급적 최단시간 내에 최소의 에너지로 해당 연산을 완료하는 설계정책을 고안하고자 한다. 이 논문은 연산과 I/O 비율이 상이한 두 알고리즘에 대해 에너지 소모특성분석, 전력효율 평가를 통해서 고효율 연산 설계법을 제시한다. 정렬 및 매트릭스 곱셈 알고리즘을 대상으로 실험하였고, 연구결과는 대규모 데이터 처리, 가공 등의 영역에서 그린 컴퓨팅을 실현하는데 기여할 것이다.

• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.127-146
• /
• 2022

Recently, as word embedding has shown excellent performance in various tasks of deep learning-based natural language processing, researches on the advancement and application of word, sentence, and document embedding are being actively conducted. Among them, cross-language transfer, which enables semantic exchange between different languages, is growing simultaneously with the development of embedding models. Academia's interests in vector alignment are growing with the expectation that it can be applied to various embedding-based analysis. In particular, vector alignment is expected to be applied to mapping between specialized domains and generalized domains. In other words, it is expected that it will be possible to map the vocabulary of specialized fields such as R&D, medicine, and law into the space of the pre-trained language model learned with huge volume of general-purpose documents, or provide a clue for mapping vocabulary between mutually different specialized fields. However, since linear-based vector alignment which has been mainly studied in academia basically assumes statistical linearity, it tends to simplify the vector space. This essentially assumes that different types of vector spaces are geometrically similar, which yields a limitation that it causes inevitable distortion in the alignment process. To overcome this limitation, we propose a deep learning-based vector alignment methodology that effectively learns the nonlinearity of data. The proposed methodology consists of sequential learning of a skip-connected autoencoder and a regression model to align the specialized word embedding expressed in each space to the general embedding space. Finally, through the inference of the two trained models, the specialized vocabulary can be aligned in the general space. To verify the performance of the proposed methodology, an experiment was performed on a total of 77,578 documents in the field of 'health care' among national R&D tasks performed from 2011 to 2020. As a result, it was confirmed that the proposed methodology showed superior performance in terms of cosine similarity compared to the existing linear vector alignment.

• Applied Biological Chemistry
• /
• v.48 no.1
• /
• pp.82-88
• /
• 2005

3D QSAR studies for the fungicidal activities against resistive phytophthora blight (RPC; 95CC7303) and sensitive phytophthora blight (Phytopthora capsici) (SPC; 95CC7105) by a series of new 2-alkoxyphenyl-3-phenylthioisoindoline-1-one derivatives (X: A=propynyl & B=2-chloropropenyl) were studied using comparative molecular field analyses (CoMFA) methodology. The CoMFA models were generated from the two different alignment, atom based fit (AF) alignment and field fit (FF) alignment. The atom based alignment exhibited a higher statistical results than that of field fit alignment. The best models, A3 and A7 using combination fields of H-bond field, standard field, LUMO and HOMO molecular orbital field as additional descriptors were selected to improve the statistic of the present CoMFA models. The statistical results of the two models showed the best predictability of the fungicidal activities based on the cross-validated value $q^2\;(r^2_{cv.}=RPC:\;0.625\;&\;SPC:\;0.834)$, non cross-validated value $(r^2_{ncv.}=RPC:\;0.894\;&\;SPC:\;0.915)$ and PRESS value (RPC: 0.105 & SPC: 0.103), respectively. Based on the findings, the predictive ability and fitness of the model for SPC was better than that of the model for RPC. The fugicidal activities exhibited a strong correlation with steric $(66.8{\sim}82.8%)$, electrostatic $(10.3{\sim}4.6%)$ and molecular orbital field (SPC: HOMO, 12.6% and RPC: LUMO, 22.9%) factors of the molecules. The novel selective character for fungicidal activity between two fungi depend on the positive charge of ortho, meta-positions on the N-phenyl ring and size of hydrophilicity of a substituents on the S-phenyl ring.

• Applied Biological Chemistry
• /
• v.48 no.3
• /
• pp.252-257
• /
• 2005

Three-dimensional quantitative structure-activity relationships (3D-QSARs) for the herbicidal activities against pre-emergence barnyard grass (Echinochloa crus-galli) by new 2-(4-(6-chloro-2-benzoxazolyloxy)phenoxy)-N-phenylpropion amide derivatives were studied quantitatively using comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA) methodologies. The best CoMFA model (AI-2) and CoMSIA model (AII-4) were derived from an atom based fit alignment and a combination of CoMFA fields. The herbicidal activities from CoMFA and CoMSIA contour maps showed that the activity will be able to be increased according to the substituents variation on the N-phenyl ring.

• The Korean Journal of Pesticide Science
• /
• v.11 no.3
• /
• pp.131-137
• /
• 2007

CoMFA and CoMSIA model were derived and reviewed on the insecticidal activities of N-substituents (X) on the imidazol ring in imidacloprid analogues at the different alignment condition. Regarding the predictability ($q^2$ or $r_{cv.}^2$) and fitness ($r_{ncv.}^2$) of the two optimized models, the atom based fit (A) alignments were better than that of the field fit (F) alignment and, on the other hand, CoMSIA (A10) model was better than CoMFA (A5) model. Also, from the most optimized CoMSIA (A10) model, the insecticidal activity by N-substituents (X) was dependence on the electrostatic field and H-bond acceptor field. It is predicted that, from the contour maps with optimized CoMSIA (A10) model, H-bond acceptors at ortho- and meta- position will contribute for improving of insecticidal activities and, as the functional groups of carbonyl oxygen atom are charged negatively and positively charged at the ortho- position of benzyl group, insecticidal activities will also be improved.