• ETRI Journal
• /
• v.43 no.2
• /
• pp.288-298
• /
• 2021

Abstractive text summarization is a process of making a summary of a given text by paraphrasing the facts of the text while keeping the meaning intact. The manmade summary generation process is laborious and time-consuming. We present here a summary generation model that is based on multilayered attentional peephole convolutional long short-term memory (MAPCoL; LSTM) in order to extract abstractive summaries of large text in an automated manner. We added the concept of attention in a peephole convolutional LSTM to improve the overall quality of a summary by giving weights to important parts of the source text during training. We evaluated the performance with regard to semantic coherence of our MAPCoL model over a popular dataset named CNN/Daily Mail, and found that MAPCoL outperformed other traditional LSTM-based models. We found improvements in the performance of MAPCoL in different internal settings when compared to state-of-the-art models of abstractive text summarization.

• Economic and Environmental Geology
• /
• v.33 no.5
• /
• pp.405-415
• /
• 2000

This study was carried out to understand the formation of acid mine drainage (AMD) by pyrophyllite (so-called Napseok)-rainwater interaction (weathering), dispersion patterns of heavy metals, and patterns of mixing with non-polluted water in the Tongnae pyrophyllite mine. Based on the mass balance and reaction path modeling, using both the geochemistry of water and occurrence of the secondary minerals (weathering products), the geochemical evolution of AMD was simulated by computer code of SOLVEQ and CHILLER. It shows that the pH of stream water is from 6.2 to 7.3 upstream of the Tongnae mine. Close to the mine, the pH decreases to 2. Despite being diluted with non-polluted tributaries, the acidity of mine drainage water maintains as far as downstream. The results of modeling of water-rock interaction show that the activity of hydrogen ion increases (pH decreases), the goncentration of ${HCO_3}^-$ decreases associated with increasing $H^+$ activity, as the reaction is processing. The concentration of ${SO_4}^{2-}$first increases minutely, but later increases rapidly as pH drops below 4.3. The concentrations of cations and heavy metals are controlled by the dissolution of reactants and re-dissolution of derived species (weathering products) according to the pH. The continuous adding of reactive minerals, namely the progressively larger degrees of water-rock interaction, causes the formation of secondary minerals in the following sequence; goethite, then Mn-oxides, then boehmite, then kaolinite, then Ca-nontronite, then Mgnontronite, and finally chalcedony. The results of reaction path modeling agree well with the field data, and offer useful information on the geochemical evolution of AMD. The results of reaction path modeling on the formation of AMD offer useful information for the estimation and the appraisal of pollution caused by water-rock interaction as geological environments. And also, the ones can be used as data for the choice of appropriate remediation technique for AMD.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2004.04a
• /
• pp.317-324
• /
• 2004

Shape design optimization of shell structure is implemented on a basis of integrated framework of geometric modeling and finite element analysis which is constructed on the geometrically exact shell theory. This shell theory enables more accurate and robust analysis for complicated shell structures, and it fits for the nature of B-spline function which Is popular modeling scheme in CAD field. Shape of laminated composite shells is optimized through genetic algorithm and sequential linear programming, because there ire numerous optima for various configurations, constraints, and searching paths. Sequential adaptation of global and local optimization makes the process more efficient. Two different optimized results of laminated composite shell structures to minimize strain energy are shown for different layup sequence.

• Applied Biological Chemistry
• /
• v.41 no.8
• /
• pp.568-571
• /
• 1998

The conformations of human insulin precursors, proinsulin and preproinsulin, are described in terms of molecular dynamics simulations. Despite the presence of the C-peptide and/or the signal peptide, molecular dynamics calculations utilizing the hydration shell model over a period of 500 ps indicate that the native conformations of the A and B chains are well conserved in both cases. These results further support the NMR spectroscopy results that the C-peptide is relatively disordered and does not influence the overall conformation of the native structure. The robustness of the native structure as demonstrated by experiment and simulation will permit future protein engineering applications, whereby the expression or purification yields can be improved upon sequence modification of the C-peptide and/or the signal peptide.

• Proceedings of the KSME Conference
• /
• 2001.06c
• /
• pp.257-262
• /
• 2001

This paper describes a specialized CAD system for injection mold design, which has been developed using the application procedure interfaces of Unigraphics. The system consists of modeling modules that are mutually independent and can be accessed without any predefined sequence. In addition, the design process modeling capability proposed in this paper facilitate mold redesign process caused by modification of part shape.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2000.10a
• /
• pp.161-164
• /
• 2000

A cyclic shop is a production system that repeatedly produces identical sets of jobs, called minimal part sets, in the same loading and processing sequence. We consider a version of cyclic shop where the operations are processed and unloaded within time limits, so called a time window. We model the shop using an event graph model, a class of Petri nets. To represent the time window constraint, we introduce places with negative time delays. From the shop modeling graph, we develop a linear system model based on the max- plus algebra and characterize the conditions on the existence of a stable schedule.

• Journal of the Korean Society of Systems Engineering
• /
• v.11 no.2
• /
• pp.75-83
• /
• 2015

This paper describes modeling & simulation(M&S) model for dynamic analysis of helicopter rotor blades. Simulation model is developed using Dymola tool which implements the open source language - Modelica. Modelica is appropriate for developing multibody dynamic analysis model. To develop an M&S model efficiently, model based systems engineering(MBSE) is applied. Some diagrams such as requirement diagram, block definition diagram and sequence diagram etc. are drawn to capture the concept of M&S model. This activity is done utilizing the open source tool - Papyrus.

• Journal of Industrial Technology
• /
• v.24 no.B
• /
• pp.139-142
• /
• 2004

To identify conserved structural or functional subunit(s) in the CP1 (editing) domains of class Ia tRNA synthetases, five available structures were compared and analyzed. Through sequence alignments of the CP1 domains, three conserved regions were found near the amino acid binding site in the editing domain. Structural overlapping of the three subunits clearly showed that there exist three common structural subunits in all of the five editing RS structures. The new alignment suggests a translocation movement of the CP1 domain caused by the binding with tRNA. Based on the experimental and modeling results, it is proposed that subunits 1 and 3 accommodate the incoming amino acid binding, while subunit 2 contributes to the interactions with the adenosine ring of the A76 to stabilize the overall tRNA binding.. Since these subunits are critical for the editing reaction, we expect that these key structures should be conserved through all class Ia editing RSs.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.51 no.4
• /
• pp.202-208
• /
• 2002

In a dynamic game where the players move in a periodical sequence, each player observes the strategy of the others. So the players who move later in a game get to know the moves of others having made before them. Those who move earlier must take this into account in devising their optimal strategy. In the Poolco model, the bidding game is executed periodically. The player participating in the bidding game accumulates the information of its own and others'strategies, and payoffs through the repeated bidding process. Thereby, the players in this game would be able to map out how get the maximum profit, and get closer to the optimal strategy. This paper presents a mathematical modeling for a player to determine his or her optimal strategy at period T, based on the information acquired from the previous rounds for the periods, T-1, T-2, and so on. The proposed modeling is demonstrated with a dynamic fame theory.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1996.04a
• /
• pp.627-630
• /
• 1996

One of the most common communication mechanisms to describe a situation or a process is a story written as an ordered sequence of events or activities. For example, a shop floor supervisor may present the operations of his manufacturing system by describing the processes of manunfacturing a product in his shop. Although IDEF3 is one of the most commonly used methods for describing a business process, but it is not common for a manufacturing process. In this study, we tried to apply IDEF3 for describing a manufacturing process. Problems and suggestions such as selection probability, programmable process modeling, manufacturing resource model were presented.