• 한국지반공학회:학술대회논문집
• /
• 한국지반공학회 2010년도 춘계 학술발표회
• /
• pp.279-290
• /
• 2010

Technical countermeasures against debris flow should be established upon the risk level of the target location. Risk of debris flow should consider the hazard imposed by debris flow and vulnerability of the facilities to debris flow. In this research, we have defined the target location for risk evaluation and suggested scoring method of hazard of debris flow and vulnerability of road to debris flow. By defining risk rank into 6 categories in terms of possibility of damage during rainfall and using the risk scores of 46 debris flow cases, we have suggested risk ranking matrix. The method can be used in ranking the drainage basin adjacent to road by simply determining the hazard with vulnerability score and can be used for planning the debris flow countermeasures.

• 말소리와 음성과학
• /
• 제7권4호
• /
• pp.9-16
• /
• 2015

In this paper, several experiments are performed according to deep neural network (DNN) based features for the performance comparison of speaker verification (SV) systems. To this end, input features for a DNN, such as mel-frequency cepstral coefficient (MFCC), linear-frequency cepstral coefficient (LFCC), and perceptual linear prediction (PLP), are first compared in a view of the SV performance. After that, the effect of a DNN training method and a structure of hidden layers of DNNs on the SV performance is investigated depending on the type of features. The performance of an SV system is then evaluated on the basis of I-vector or probabilistic linear discriminant analysis (PLDA) scoring method. It is shown from SV experiments that a tandem feature of DNN bottleneck feature and MFCC feature gives the best performance when DNNs are configured using a rectangular type of hidden layers and trained with a supervised training method.

• The Journal of the Acoustical Society of Korea
• /
• 제16권3E호
• /
• pp.19-23
• /
• 1997

The goal of postprocessing in automatic speech recognition is to improve recognition performance by utterance verification at the output of recognition stage. It is focused on the effective rejection of out-of vocabulary words based on the confidence score of hypothesized candidate word. We present two methods for computing confidence scores. Both methods are based on the distance between each observation vector and the representative code vector, which is defined by the most likely code vector at each state. While the first method employs simple time normalization, the second one uses a normalization technique based on the concept of on-line garbage mode[1]. According to the speaker independent isolated words recognition experiment with discrete density HMM, the second method outperforms both the first one and conventional likelihood ratio scoring method[2].

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제10권3호
• /
• pp.1071-1090
• /
• 2016

Most of scheduling methods in the Grid only consider one special attribute of the resource or one aspect of QoS (Quality of Service) of the job. In this paper, we focus on the problem that how to consider two aspects simultaneously. Based on the requirements of the jobs and the attributes of the resources, jobs are categorized into three kinds: CPU-overload, memory-overload, and bandwidth-overload jobs. One job may belong to different kinds according to different attributes. We schedule the jobs in different categories in different orders, and then propose a scheduling method-MTS (multiple attributes scheduling method) to schedule Grid resources. Based on the comparisons between our method, Min-min, ASJS (Adaptive Scoring Job Scheduling), and MRS (Multi-dimensional Scheduling) show: (1) MTS reduces the execution time more than 15% to other methods, (2) MTS improves the number of the finished jobs before the deadlines of the jobs, and (3) MTS enhances the file size of transmitted files (input files and output files) and improves the number of the instructions of the finished jobs.

• Journal of Information Processing Systems
• /
• 제12권3호
• /
• pp.455-467
• /
• 2016

This paper presents the constituent-based approach for aligning bilingual multiword expressions, such as noun phrases, by considering the relationship not only between source expressions and their target translation equivalents but also between the expressions and constituents of the target equivalents. We only considered the compositional preferences of multiword expressions and not their idiomatic usages because our multiword identification method focuses on their collocational or compositional preferences. In our experimental results, the constituent-based approach showed much better performances than the general method for extracting bilingual multiword expressions. For our future work, we will examine the scoring method of the constituent-based approach in regards to having the best performance. Moreover, we will extend target entries in the evaluation dictionaries by considering their synonyms.

• 디지털산업정보학회논문지
• /
• 제7권2호
• /
• pp.19-27
• /
• 2011

Recent development of information technologies produces a lot of community services. Social Network Service is one of the community services on the world wide webs. In the Social Network Service, a user can register other users as friends and enjoy communication through a virtual message. Previous researches show a few social service methods using manually generated tagging. However, the manual social tagging is not widely used in many social network services. Moreover, they do not consider ubiquitous computing environment. We propose a recommendation method for social service using contexts in ubiquitous environment. Our method scores documents based on context tags and social network services. Our social scoring model is computed by both a tagging score of a document and a tagging score of a document that was tagged by a user's friends.

• 산업경영시스템학회지
• /
• 제14권23호
• /
• pp.57-64
• /
• 1991

The preliminary, advanced and final evaluation of R&D project is essential in order to reduce loss of resource and opportunity and to minimize uncertainty resulting from optimal selection and efficient progress of R&D project. This thesis examined characteristics of deterministic evaluation, economical evaluation, and OR-approach evaluation as theoretical methodology of evaluation of R&D project applicable to food industry in Korea by using scoring method, one of deterministic evaluations. In addition, this thesis divided the evaluation factors for preliminary evaluation of R&D project into 5 groups and 30 factors on basis of the environment of domestic companies and set up the standard of each evaluation factors and contains marking-selecting way. But, generally, the evaluation model by this thesis, as the conditions of the business company environment are different to each other, contents to be set up evaluation factors, evaluation standard and decision method conforming to each the environment of the business companies with referring to as one standard of evaluation project for selecting R&D project.

• 정신신체의학
• /
• 제5권1호
• /
• pp.52-62
• /
• 1997

This study was designed to determine the frequency of premenstrual dysphoric disorder in gynecological outpatients, and also attempted to compare premenstrual change characteristics, functional impairment due to premenstrual changes and frequency of risk factors reported by women with confirmed premenstrual changes$(PMC^+)$(n=17) and those without confirmed premenstrual changes$(PMC^-)$(n=23). Forty gynecological outpatients who complained of premenstrual discomforts were asked to complete questionnaires on menstrual history, obstetric-gynecological history, and premenstrual change and functional impairment. The women were also asked to complete a daily rating form based on DSM-IV diagnostic criteria for one menstrual cycle. Absolute severity method, effect size method and percent change method were used to assess changes between follicular phase and luteal phase. The results of the study were as follows: 1) The frequency of premenstrual dysphoric disorder according to each of the three methods was 5% for the absolute severity method, 15% for the effect size method, and 27.5% for the percent change method. 2) The frequently reported symptoms were as follow: physical symptoms(64.7%) : lethargy, easy fatigability, or marked lack of energy(41.2%) : decreased interest in usual activities(29.4%) ; and marked affective lability(23.5%). 3) There were no significant differences in onset ages of premenstrual changes, regularities of premenstrual changes and changes of severity and duration of premenstrual symptoms over time between women with and without confirmed premenstrual changes. However, women with confirmed premenstrual changes reported both physical and emotional symptoms as earliest symptoms most frequently, while women without confirmed premenstrual changes reported only physical symptoms most frequently. 4) functional impairment was significantly higher in women with confirmed premenstrual changes than those without confirmed premenstrual changes, but impairment was not severe. 5) No differences were found between women with and without confirmed premenstrual changes in risk factors including demographic data, menstrual and obstetric and gynecological history. These results suggest that the prevalence of premenstrual dysphoric disorder varies with scoring methods. The women with confirmed premenstrual changes reported physical symptoms most frequently(64.7%). functional impairment was significantly higher in women with confirmed premenstrual changes, but impairment was not severe.

• ETRI Journal
• /
• 제42권5호
• /
• pp.761-772
• /
• 2020

This paper presents an automatic proficiency assessment method for a non-native Korean read utterance using bidirectional long short-term memory (BLSTM)-based acoustic models (AMs) and speech data augmentation techniques. Specifically, the proposed method considers two scenarios, with and without prompted text. The proposed method with the prompted text performs (a) a speech feature extraction step, (b) a forced-alignment step using a native AM and non-native AM, and (c) a linear regression-based proficiency scoring step for the five proficiency scores. Meanwhile, the proposed method without the prompted text additionally performs Korean speech recognition and a subword un-segmentation for the missing text. The experimental results indicate that the proposed method with prompted text improves the performance for all scores when compared to a method employing conventional AMs. In addition, the proposed method without the prompted text has a fluency score performance comparable to that of the method with prompted text.

• 분석과학
• /
• 제34권2호
• /
• pp.56-67
• /
• 2021

Due to the difficulty of grading visualized fingerprints with previously known evaluation methods for the comparison of split fingerprints, a preliminary study was performed with the densitometric image analysis (DIA) method as a potential quantitative and supplementary evaluation method. Each image of inked split fingerprints was divided into 4 zones for analysis. Weekly intra- and inter- analysis by two analysts with three whole fingerprints that were constructed by combining inked split fingerprints showed that the average area values and the ranges of difference fluctuation were not significantly different between strong fingerprints and strong-weak pairs, while they were different in weak fingerprints and weak-weak pairs. In the case of weak fingerprints, the exact acquisition of ridges was difficult and this seemed to influence the results. An additional study is needed for the improved reliability using DIA method with weak fingerprints such as 8 zones division rather than 4 zones. In addition, the analysis results performed by several analysts at different times should be used to improve the reliability of the analysis method further. Based on the above result, it can be judged that utilizing the DIA method as a secondary evaluation method of the existing scoring system would be effective with the additional studies especially on weak fingerprints.