• Annual Conference on Human and Language Technology
• /
• 2016.10a
• /
• pp.111-116
• /
• 2016

의미역은 자연어 문장의 서술어와 관련된 논항의 역할을 설명하는 것으로, 주어진 서술어에 대한 논항 인식(Argument Identification) 및 분류(Argument Labeling)의 과정을 거쳐 의미역 결정(Semantic Role Labeling)이 이루어진다. 이를 위해서는 격틀 사전을 이용한 방법이나 말뭉치를 이용한 지도 학습(Supervised Learning) 방법이 주를 이루고 있다. 이때, 격틀 사전 또는 의미역 주석 정보가 부착된 말뭉치를 구축하는 것은 필수적이지만, 이러한 노력을 최소화하기 위해 본 논문에서는 비모수적 베이지안 모델(Nonparametric Bayesian Model)을 기반으로 서술어에 가능한 의미역을 추론하는 비지도 학습(Unsupervised Learning)을 수행한다.

• The Korean Journal of Applied Statistics
• /
• v.27 no.6
• /
• pp.867-889
• /
• 2014

Nonparametric Bayesian (np Bayes) statistical models are popularly used in a variety of research areas because of their flexibility and computational convenience. This paper reviews the np Bayes models focusing on biomedical research applications. We review key probability models for np Bayes inference while illustrating how each of the models is used to answer different types of research questions using biomedical examples. The examples are chosen to highlight the problems that are challenging for standard parametric inference but can be solved using nonparametric inference. We discuss np Bayes inference in four topics: (1) density estimation, (2) clustering, (3) random effects distribution, and (4) regression.

• The KIPS Transactions:PartD
• /
• v.9D no.3
• /
• pp.389-398
• /
• 2002

Software reliability growth models are used in testing stages of software development to model the error content and time intervals between software failures. This paper presents a stochastic model for the software failure phenomenon based on a nonhomogeneous Poisson process(NHPP) and performs Bayesian inference using prior information. The failure process is analyzed to develop a suitable mean value function for the NHPP ; expressions are given for several performance measure. Actual software failure data are compared with several model on the constant reflecting the quality of testing. The performance measures and parametric inferences of the suggested models using Rayleigh distribution and Laplace distribution are discussed. The results of the suggested models are applied to real software failure data and compared with Goel model. Tools of parameter point inference and 95% credible intereval was used method of Gibbs sampling. In this paper, model selection using the sum of the squared errors was employed. The numerical example by NTDS data was illustrated.

• The Korean Journal of Applied Statistics
• /
• v.34 no.3
• /
• pp.477-489
• /
• 2021

COVID-19 has spread seriously around the world in 2020 and it is still significantly affecting our whole daily life. Currently, the whole world is still undergoing the pandemic and South Korea is no exception to it. During the pandemic, South Korea had several events that prevented or accelerated its spread. To establish the prevention policies for infectious diseases, it is very important to evaluate the intervention effect of such events. The susceptible-infected-removed (SIR) model is often used to describe the dynamic behavior of the spread of infectious diseases through ordinary differential equations. However, the SIR model is a deterministic model without considering the uncertainty of observed data. To consider the uncertainty in the SIR model, the Bayesian approach can be employed, and this approach allows us to evaluate the intervention effects by time-varying functions of the infection rate in the SIR model. In this study, we describe the time trend of the spread of COVID-19 in South Korea and investigate the intervention effects for the events using the stochastic SIR model based on the Bayesian approach.

• The Korean Journal of Applied Statistics
• /
• v.6 no.1
• /
• pp.53-66
• /
• 1993

A Bayesian approach is suggested to the multi-proportions randomized response model. O'Hagan's (1987) Bayes linear estimator is extended to the inference of unrelated question-type randomized response model. Also some numerical comparisons are provided to show the performance of the Bayes linear estimator under the Dirichlet prior.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.1
• /
• pp.16-21
• /
• 2001

전자 상거래에서 고객의 성향을 이해하기 위해서는 일반적으로 판매 실무에서의 경험과 전문적인 지식을 필요로 하게 된다. 데이터 마이닝은 고객들에 대한 데이터의 분석을 통해서 이러한 성향들을 알아내는 것을 목표로 한다. 베이지안 네트워크는 DAG(Directed Acyclic Graph)를 이용하여 데이터의 구조를 시각적으로 표현하여 주는 확률모형으로 변수사이의 종속관계를 밝히고 데이터 마이닝의 기법으로 이용할 수 있다. 본 논문에서는 베이지안 네트워크를 사용하여 전자 상거래 고객들의 성향을 분석하기 위한 방법을 제시한다. 또한 고객성향에 대한 주요 요인을 분석하기 위해 Discriminant 모형을 이용하고 그 유용성을 다른 방법들과 비교하였다.

• Journal of Korea Water Resources Association
• /
• v.32 no.1
• /
• pp.15-29
• /
• 1999

A new methodology to analyze and quantify regional meteorological drought based on annual precipitation data has been introduced in this paper In this study, based on posterior probability estimator and Bayesian classifier in Spatial Analysis Neural Network (SANN), point drought probabilities categorized as extreme, severe, mild, and non drought events has been defined, and a Bayesian Drought Severity Index (BPSI) has been introduced to classify the region of interest into four drought severities. In addition, to estimate the regional drought severity for the entire region, regional extreme, severe, mild, and non drought probabilities which are the areal averages of point drought probabilities over the region has been computed and applied. In this study, the proposed methodology has been applied to analyze the regional drought of South Korea during 1967-1996 years. The drought severity for the whole South Korea was defined spatially at each year and each year was classified in a drought severity criterion. The results may be useful for water manager to understand the South Korean drought with respect to the spatial and temporal variation.

• Journal of KIISE
• /
• v.44 no.10
• /
• pp.1026-1033
• /
• 2017

In this paper a Bayesian modeling and duration-based prediction method is proposed for health clinic time series data using the Hierarchical Dirichlet Process Hidden Markov Model (HDP-HMM). HDP-HMM is a Bayesian extension of HMM which can find the optimal number of health states, a number which is highly uncertain and even difficult to estimate under the context of health dynamics. Test results of HDP-HMM using simulated data and real health clinic data have shown interesting modeling behaviors and promising prediction performance over the span of up to five years. The future of health change is uncertain and its prediction is inherently difficult, but experimental results on health clinic data suggests that practical long-term prediction is possible and can be made useful if we present multiple hypotheses given dynamic contexts as defined by HMM states.

• The Korean Journal of Applied Statistics
• /
• v.31 no.5
• /
• pp.555-572
• /
• 2018

Noise reduction processes that reduce or eliminate noise (caused by a variety of reasons) in noise contaminated image is an important theme in image processing fields. Many studies are being conducted on noise removal processes due to the importance of distinguishing between noise added to a pure image and the unique characteristics of original images. Adaptive filter and sigma filter are typical noise reduction filters used to reduce or eliminate noise; however, their effectiveness is affected by accurate noise estimation. This study generates a distribution of noise contaminating image based on a Dirichlet normal mixture model and presents a Bayesian approach to distinguish the characteristics of an image against the noise. In particular, to distinguish the distribution of noise from the distribution of characteristics, we suggest algorithms to develop a Bayesian inference and remove noise included in an image.

• Communications for Statistical Applications and Methods
• /
• v.15 no.4
• /
• pp.531-542
• /
• 2008

The unit root test is the major tool for determining whether we use differencing or detrending to eliminate the trend from time series data. Dickey-Fuller test (Dickey and Fuller, 1979) has the low power of test when the sample size is small or the true coefficient of AR(1) process is almost unit root and the Bayesian unit root test has complicated testing procedure. We propose a new unit root testing procedure, which mixed Bayesian approach with the traditional testing procedure. Using simulation studies, our approach showed locally higher powers than Dickey-Fuller test when the sample size is small or the time series has almost unit root and simpler procedure than Bayesian unit root test procedure. Proposed testing procedure can be applied to the time series data that are not observed as process with unit root.