• ETRI Journal
• /
• 제46권1호
• /
• pp.35-47
• /
• 2024

This study introduces CR-M-SpanBERT, a coreference resolution (CR) model that utilizes multiple embedding-based span bidirectional encoder representations from transformers, for antecedent recognition in natural language (NL) text. Information extraction studies aimed to extract knowledge from NL text autonomously and cost-effectively. However, the extracted information may not represent knowledge accurately owing to the presence of ambiguous entities. Therefore, we propose a CR model that identifies mentions referring to the same entity in NL text. In the case of CR, it is necessary to understand both the syntax and semantics of the NL text simultaneously. Therefore, multiple embeddings are generated for CR, which can include syntactic and semantic information for each word. We evaluate the effectiveness of CR-M-SpanBERT by comparing it to a model that uses SpanBERT as the language model in CR studies. The results demonstrate that our proposed deep neural network model achieves high-recognition accuracy for extracting antecedents from NL text. Additionally, it requires fewer epochs to achieve an average F1 accuracy greater than 75% compared with the conventional SpanBERT approach.

• 한국음향학회지
• /
• 제23권5호
• /
• pp.390-399
• /
• 2004

본 논문에서는 음성 웹 언어인 VoiceXML의 작성 효율을 향상하기 위한 다중 뷰 편집기를 설계 및 구현하였다. VoiceXML 다중 뷰 편집기는 다중 뷰를 제공하기 위해 MVC (Model-View-Controller) 프레임워크을 이용하였다. MVC 프레임워크를 이용한 다중 뷰 편집기는 핵심 자료구조인 모델과 인터페이스인 뷰, 모델과 뷰를 제어하기 위한 제어기로 구성된다. MVC 프레임워크에서 모델은 추상 구문 트리와 추상 문법으로 구성되며 뷰는 역파싱 규칙과 역파서로 구성되고 제어기는 명령어 처리기와 트리 조작기로 구성된다. VoiceXML 다중 뷰 편집기는 문서의 구조, 내용, 흐름을 동시에 보여주어 기존 XML 편집기의 단점을 극복할 수 있다. MVC 프레임워크가 적용된 VoiceXML 다중 뷰 편집기는 여러 편집기를 통해 동시에 다양한 편집 뷰 (View)를 제공함으로써 사용자에게 음성 웹 문서 작성의 편의성을 제공하여 효율을 높일 수 있으며 여러 개의 뷰가 하나의 모델을 가짐으로써 편집기들의 무결성을 보장하도록 하였다.

• 한국산업정보학회논문지
• /
• 제16권4호
• /
• pp.99-105
• /
• 2011

본 논문은 다물체 복합시스템 인자의 변위에 강인한 중앙집중식 연속반복학습제어기(Repetitive Controller, RC)를 설계하는 방법을 소개하고자 한다. 이때 사용되는 불확실 인자들은 확률분포함수에 의해 무작위로 설정되게 된다. 분포함수를 직접 적용하는 대신, 본 제어기는 설정된 확률함수로부터 생성된 모형을 기본으로 설계하였다. 이러한 중앙집중식 복수모형 설계기법으로 엄의의 분포함수로 구성된 수많은 불확실 인자들을 다룰 수 있다. 그러므로, 제어기는 반복영역에서 수렴성을 보장하는 비용함수를 주파수영역에서 최소화함으로써 유도할 수 있다. 다물체 복합시스템에서 중앙집중식 복수모형설계 기법을 단수모형 설계기법과 함께 제안하였다.

• Ocean and Polar Research
• /
• 제34권2호
• /
• pp.185-195
• /
• 2012

This paper focused on forecasting a short-term production of oysters, which have been farmed in Korea, with distinct periodicity of production by year, and different production level by month. To forecast a short-term oyster production, this paper uses monthly data (260 observations) from January 1990 to August 2011, and also adopts several econometrics methods, such as Multiple Regression Analysis Model (MRAM), Seasonal Autoregressive Integrated Moving Average (SARIMA) Model, and Vector Error Correction Model (VECM). As a result, first, the amount of short-term oyster production forecasted by the multiple regression analysis model was 1,337 ton with prediction error of 246 ton. Secondly, the amount of oyster production of the SARIMA I and II models was forecasted as 12,423 ton and 12,442 ton with prediction error of 11,404 ton and 11,423 ton, respectively. Thirdly, the amount of oyster production based on the VECM was estimated as 10,425 ton with prediction errors of 9,406 ton. In conclusion, based on Theil inequality coefficient criterion, short-term prediction of oyster by the VECM exhibited a better fit than ones by the SARIMA I and II models and Multiple Regression Analysis Model.

• 한국음향학회지
• /
• 제21권4호
• /
• pp.387-393
• /
• 2002

비정상 잡음에 오염된 음성신호의 향상을 위하여 혼합 은닉필터모델 (HFM: Hidden Filter Model)에 기초한 기법을 제안하였다. 오염된 음성신호를 선형상태방정식으로 모델링하고 파라미터는 마코프 모델에 따른다고 가정하였다. 이 파라미터들은 잡음에 오염되지 않은 학습신호로부터 추정할 수 있다. 추정과정은 혼합 상호복합모델 (IMM: Interacting Multiple Model)에 기초하여 이루어지며, 음성신호의 추정값은 상호작용하는 병렬의 칼만 필터들의 가중합으로 주어진다. 실험결과로부터 제안한 방법의 성능이 기존의 방법에 비해 개선되었음을 확인할 수 있었다.

• 한국환경과학회지
• /
• 제14권11호
• /
• pp.1049-1056
• /
• 2005

[ $NO_2$ ] concentration characteristics of Busan metropolitan city was analysed by statistical method using hourly $NO_2$ concentration data$(1998\~2000)$ collected from air quality monitoring sites of the metropolitan city. 4 representative regions were selected among air quality monitoring sites of Ministry of environment. Concentration data of $NO_2$, 5 air pollutants, and data collected at AWS was used. Both Stepwise Multiple Regression model and ARIMA model for prediction of $NO_2$ concentrations were adopted, and then their results were compared with observed concentration. While ARIMA model was useful for the prediction of daily variation of the concentration, it was not satisfactory for the prediction of both rapid variation and seasonal variation of the concentration. Multiple Regression model was better estimated than ARIMA model for prediction of $NO_2$ concentration.

• 제어로봇시스템학회논문지
• /
• 제15권9호
• /
• pp.939-945
• /
• 2009

In this paper, we proposed a method which corresponds people under the structured spaces with multiple cameras. The correspondence takes an important role for using multiple camera system. For solving this correspondence, the proposed method consists of three main steps. Firstly, moving objects are detected by background subtraction using a multiple background model. The temporal difference is simultaneously used to reduce a noise in the temporal change. When more than two people are detected, those detected regions are divided into each label to represent an individual person. Secondly, the detected region is segmented as features for correspondence by a criterion with the color distribution and context information of human body. The segmented region is represented as a set of blobs. Each blob is described as Gaussian probability distribution, i.e., a person model is generated from the blobs as a Gaussian Mixture Model (GMM). Finally, a GMM of each person from a camera is matched with the model of other people from different cameras by maximum likelihood. From those results, we identify a same person in different view. The experiment was performed according to three scenarios and verified the performance in qualitative and quantitative results.

• Asia Marketing Journal
• /
• 제18권4호
• /
• pp.125-138
• /
• 2017

The goal of the study is to understand how consumers' constraint as opposed to utility structure gives rise to final decision when consumers purchase more than one variant of product at a time, i.e., horizontal variety seeking or multiple-discreteness. Purchase and consumption decision not only produces utility but also involves some sort of cognitive pressure. Past consumption or last purchase is likely to be linked to this burden we face such as concern for obesity, risk of harm, and guilt for mischief. In this research, the existence and the role of dynamic constraint are investigated through a microeconomic utility model with multiple dynamic constraint. The model is applied to the salty snacks data collected from field study where burden for spiciness serves as a constraint. The results are compared to the conventional multiple discreteness choice models of static constraints, and policy implications on price discounts is explored. The major findings are that first, one would underestimate the level of consumer preference for product offerings when ignoring the carry-over of the concern from the past consumption, and second, the impact of price promotion on demand would be properly evaluated when the model allows for the role of constraint as both multiple and dynamic. The current study is different from the existing studies in two ways. First, it captures the effect of 'mental constraint' on demand in formal economic model. Second, unlike the state dependence well documented in the literature, the study proposes the notion of state dependence in different way, via constraint rather than utility.

• Journal of the Korean Data and Information Science Society
• /
• 제9권2호
• /
• pp.165-172
• /
• 1998

This paper considers a linear regression model with censored data where each error term follows a multivariate normal distribution. In this paper we consider the diffuse prior distribution for parameters of the linear regression model. With censored data we derive the full conditional densities for parameters of a multiple regression model in order to obtain the marginal posterior densities of the relevant parameters through the Gibbs Sampler, which was proposed by Geman and Geman(1984) and utilized by Gelfand and Smith(1990) with statistical viewpoint.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2005년도 학술발표회(1)
• /
• pp.57-58
• /
• 2005