• Proceedings of the Materials Research Society of Korea Conference
• /
• 1998.08a
• /
• pp.50.2-50
• /
• 1998

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.5
• /
• pp.662-667
• /
• 2022

Semantic role labeling task(SRL) is to extract predicate and arguments such as agent, patient, place, time. In the previously SRL task studies, a pipeline method extracting linguistic features of sentence has been proposed, but in this method, errors of each extraction work in the pipeline affect semantic role labeling performance. Therefore, methods using End-to-End neural network model have recently been proposed. In this paper, we propose a neural network model using the Biaffine Average Attention model for SRL task. The proposed model consists of a structure that can focus on the entire sentence information regardless of the distance between the predicate in the sentence and the arguments, instead of LSTM model that uses the surrounding information for prediction of a specific token proposed in the previous studies. For evaluation, we used F1 scores to compare two models based BERT model that proposed in existing studies using F1 scores, and found that 76.21% performance was higher than comparison models.

• Journal of KIISE
• /
• v.42 no.4
• /
• pp.475-482
• /
• 2015

Developing a high-performance Semantic Role Labeling (SRL) system for a domain requires manually annotated training data of large size in the same domain. However, such SRL training data of sufficient size is available only for a few domains. Performances of Korean SRL are degraded by almost 15% or more, when it is directly applied to another domain with relatively small training data. This paper proposes two techniques to minimize performance degradation in the domain transfer. First, a domain adaptation algorithm for Korean SRL is proposed which is based on the prior model that is one of domain adaptation paradigms. Secondly, we proposed to use simplified features related to morphological and syntactic tags, when using small-sized target domain data to suppress the problem of data sparseness. Other domain adaptation techniques were experimentally compared to our techniques in this paper, where news and Wikipedia were used as the sources and target domains, respectively. It was observed that the highest performance is achieved when our two techniques were applied together. In our system's performance, F1 score of 64.3% was considered to be 2.4~3.1% higher than the methods from other research.

• Proceedings of the Materials Research Society of Korea Conference
• /
• 1998.08a
• /
• pp.50.1-50
• /
• 1998

• Korean Journal of Soil Science and Fertilizer
• /
• v.50 no.3
• /
• pp.179-186
• /
• 2017

This study was conducted to identify characteristics of domestic pine forest soils and to elucidate major soil influencing factors for natural regeneration. We analyzed the physico-chemical characteristics of the soil samples collected from 23 pine forests and confirmed the similar results with the forest soil characteristics. Soil pH, organic matter content, total nitrogen, exchangeable Ca, silt content, and exchangeable Al were selected as the major soil factors among the exposed soils through 10 days of pine seedlings exposure and cultivation experiments and statistical analysis. Multiple regression analysis showed that soil pH had a positive effect on specific root length (SRL) of red pine seedlings and exchangeable Al was a significant factor affecting negative change in SRL. Taken together, the reduction of exchangeable Al by soil pH adjustment would be helpful for natural regeneration by restoring the forest and improving the fine root and root integrity of pine seedlings. Therefore, soil pH and exchangeable Al could be recommended as a major soil factor to be carefully considered in the monitoring and management of soil in pine forests that need to be renewed in the future.

• Journal of KIISE
• /
• v.42 no.2
• /
• pp.220-226
• /
• 2015

Semantic role labeling (SRL) systems determine the semantic role labels of the arguments of predicates in natural language text. An SRL system usually needs to perform four tasks in sequence: Predicate Identification (PI), Predicate Classification (PC), Argument Identification (AI), and Argument Classification (AC). In this paper, we use the Korean Propbank to develop our Korean semantic role labeling system. We describe our Korean semantic role labeling system that uses sequence labeling with structured Support Vector Machine (SVM). The results of our experiments on the Korean Propbank dataset reveal that our method obtains a 97.13% F1 score on Predicate Identification and Classification (PIC), and a 76.96% F1 score on Argument Identification and Classification (AIC).

• Nuclear Engineering and Technology
• /
• v.46 no.5
• /
• pp.595-604
• /
• 2014

SILER (Seismic-Initiated event risk mitigation in LEad-cooled Reactors) is a Collaborative Project, partially funded by the European Commission in the $7^{th}$ Framework Programme, aimed at studying the risk associated to seismic-initiated events in Generation IV Heavy Liquid Metal reactors, and developing adequate protection measures. The project started in October 2011, and will run for a duration of three years. The attention of SILER is focused on the evaluation of the effects of earthquakes, with particular regards to beyond-design seismic events, and to the identification of mitigation strategies, acting both on structures and components design. Special efforts are devoted to the development of seismic isolation devices and related interface components. Two reference designs, at the state of development available at the beginning of the project and coming from the $6^{th}$ Framework Programme, have been considered: ELSY (European Lead Fast Reactor) for the Lead Fast Reactors (LFR), and MYRRHA (Multi-purpose hYbrid Research Reactor for High-tech Applications) for the Accelerator-Driven Systems (ADS). This paper describes the main activities and results obtained so far, paying particular attention to the development of seismic isolators, and the interface components which must be installed between the isolated reactor building and the non-isolated parts of the plant, such as the pipe expansion joints and the joint-cover of the seismic gap.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2001.04a
• /
• pp.193-196
• /
• 2001

본 논문은 ATM(Abstract Timed Machine)으로 명세된 실시간 시스템에 대한 재/역공학 측면에서의 개발 및 검증을 위한 코드 변환기를 설계한다. ATM은 모드(mede), 전이(transition), 포트(per)로 구성되는데, 순공학 과정에서 실시간 시스템을 설계, 명세 하는 기존의 정형기법과는 달리 ATM은 소프트웨어의 순환공학 과정에서 사용하기 위해 설계되었다. ATM은 기존 정형기법이 순공학 과정에서의 특정 물리적 환경에서 실행되는 동적행위에 대한 부적절한 표현에 대해 순환공학에서 실시간 시스템의 속성은 물론 특정 환경과 동적 정보 등을 명세하기 위한 정형 기법으로서, 본 논문에서는 DoME을 이용하여 ATM 명세도구를 개발하고 이를 이용하여 실시간 시스템의 특정 요구사항을 위한 ATM을 명세한다. 또한 해당 ATM을 DOME/ATM 스크립트 파일로 저장하고 이에 대한 명세분석을 통해 노드와 관련된 정보를 추출하여 다른 분석도구가 이용할 수 있도록 DB에 저장하거나 매개 언어인 SRL/ATM으로 변환하며, 이러한 SRL/ATM으로부터 실행코드에 대한 관련 정보를 추출하여 실시간 시스템 개발 및 검증을 위한 Ada 코드를 생성할 수 있는 코드 변환기를 설계한다.

• Journal of Nuclear Fuel Cycle and Waste Technology(JNFCWT)
• /
• v.4 no.3
• /
• pp.227-234
• /
• 2006

Two candidate glasses, AG8W1 and DG-2, have been developed for vitrifying the mixture of low activity resin, zeolite and Dry Active Waste(DAW), and DAW solely, respectively. In order to evaluate the chemical durability of the glasses, two different leaching tests, Product Consistency Test(PCT) and Vapor Hydration Test(VHT), have been performed. As the results of PCT performed from 7 to 120 days, the leach rates of B, Na, Si and Li in the glasses were much lower than those of the benchmark glass(SRL-EA). As the result of VHT peformed for 7 days, the leach rates were 2 and $10g/m^2/day$ for AG8W1 and DG-2, respectively, The results of VHT met the regulatory guideline( $<50g/m^2/day$) for the low activity glasses of Hanford in the USA. Consequently, two candidate glasses to be used at a commercial operation in the future showed that their chemical durability is satisfactory according to the results of two leaching tests.

• Asian-Australasian Journal of Animal Sciences
• /
• v.16 no.6
• /
• pp.852-862
• /
• 2003

An experiment was conducted to study the effect of ionophore enriched cold processed mineral block supplemented with urea molasses on microbial growth and rumen fermentation. Twelve adult male crossbred cattle were divided into four groups on body weight basis. Animals were given wheat straw as a basal diet. The animals of group I and II were supplemented with concentrate mixture and animals of group III and IV were supplemented with cold processed urea molasses mineral block (UMMB). Thirty mg monensin/day/animal were supplemented to the animals of group II and 35 ppm monensin were incorporated in the UMMB supplemented to the animals of group IV. Dry matter (DM) intake did not differ significantly among groups. Mean rumen pH was higher in UMMB fed animals. Total volatile fatty acids (TVFA) concentration (mmole/L strained rumen liquor (SRL) in group III (113.19) was significantly (p<0.05) higher than those of group I (105.83) and II (108.74) but similar to group IV (109.34). TVFA production (mole/day) was similar in all the groups. The molar proportion of acetate was significantly (p<0.01) higher in the group I (59.56) than those of group II (51.73) and IV (55.91) but similar to group III (57.12). The molar proportion of propionate was significantly (p<0.01) higher in the monensin treated groups i.e. group II (38.38) and IV (36.26) than those of group I (27.78) and III (33.06). Butyrate molar percent was significantly (p<0.01) higher in group I (12.65) than those of group II (10.19), group III (9.83) and IV (7.84). The reduction of acetate and butyrate was due to UMMB and monensin resulted in lower A:P ratio. Average bacterial pool and bacterial production rate did not differ significantly among groups. Total N concentration (mg/100 ml SRL) was significantly (p<0.01) higher in the group I (55.30) and III (57.70) as compared to the group II (47.97) and IV (47.59). Ammonia-N concentration (mg/100 ml SRL) of group III (34.99) was significantly (p<0.01) higher than that of the group I (25.76) which was again significantly (p<0.01) higher than that of the group II (20.79) and IV (19.83) indicating slower release of ammonia due to monensin in diet. Total bacterial, cellulolytic, proteolytic bacterial and fungal count at 4 h post feeding did not differ significantly (p<0.05) among treatment groups. However, methanogenic bacterial count was significantly (p<0.01) higher in the group I (11.80) compared to the group II (8.43) which was significantly (p<0.01) higher than that of the group III (4.70) and IV (2.90). Average protozoal population was affected by both treatments. Thus feeding of UMMB and monensin in diet affected the rumen fermentation pattern towards propionate production, slower release of ammonia and reduction in methanogenic bacteria in the rumen.