Multitasking and Information Gain: Effects of Relevance between Tasks

황유리,정세훈 (주)에스비에스 2018 미디어경제와 문화 Vol.16 No.2

This study examined the role of task relevance in multitasking effects. Based on the capacity model of attention (Kahneman, 1973) and the cognitive dimensional framework of media multitasking (Wang, Irwin, Cooper, & Srivastava., 2015), we predicted that relevant multitasking (i.e., a multitasking combination composed of two relevant tasks) would result in greater information gain compared to irrelevant multitasking (i.e., a multitasking combination composed of two irrelevant tasks). Study 1 compared three conditions: non-multitasking (video newscast only), irrelevant multitasking (video newscast and unrelated text news), and relevant multitasking (video newscast and related text news). The results showed that the relevant multitasking did not reduce information gain (compared to the non-multitasking condition), whereas irrelevant multitasking did. Study 2 compared two conditions: irrelevant multitasking (video newscast and text news in a different order) and relevant multitasking (video newscast and text news in the same order). Study 2 also found that relevant multitasking resulted in greater information gain than irrelevant multitasking. The results of two studies suggest the beneficial effect of relevant multitasking.

When does multitasking facilitate information processing? Effects of Internet-based multitasking on information seeking and information gain

황유리,정세훈 한국사회과학협의회 2014 Korean Social Science Journal Vol.41 No.2

This study examined whether Internet-based multitasking facilitates informationgain by allowing users to seek additional information online. Study 1, using survey data,suggested that TV-Internet multitasking increased information gain, whereas TV-print mediamultitasking reduced it. In addition, online information seeking mediated the effect of TV-Internet multitasking on information gain. Study 2, using experimental data, confirmed thedifferential effects of TV-Internet multitasking and TV-print multitasking on informationgain. The theoretical and practical implications of these findings are further discussed.

A Personalized Approach for Recommending Useful Product Reviews Based on Information Gain

( Joon Yeon Choeh ),( Hong Joo Lee ),( Sung Joo Park ) 한국인터넷정보학회 2015 KSII Transactions on Internet and Information Syst Vol.9 No.5

Customer product reviews have become great influencers of purchase decision making. To assist potential customers, online stores provide various ways to sort customer reviews. Different methods have been developed to identify and recommend useful reviews to customers, primarily using feedback provided by customers about the helpfulness of reviews. Most of the methods consider the preferences of all users to determine whether reviews are helpful, and all users receive the same recommendations. In this paper, we assessed methods for generating personalized recommendations based on information gain. The information gain approach was extended to consider each individual`s preference together with votes of other users. A total of 172 respondents rated 48 reviews selected from Amazon.com using a 7-point Likert scale. The performance of the devised methods was measured by varying the ratio of training sets and number of recommendations for the data collected. The personalized methods outperformed the existing information gain method, which takes into account the votes from all users. The greatest precision was achieved by the personalized method and a method employing selective use of predictions from the personalized method combined with the existing method based on all users` reviews. However, the personalized method, which classified helpful reviews based on each user`s threshold value, showed statistically better performance.

문장 분류를 위한 정보 이득 및 유사도에 따른 단어 제거와 선택적 단어 임베딩 방안

이민석,양석우,이홍주 한국지능정보시스템학회 2019 지능정보연구 Vol.25 No.4

Dimensionality reduction is one of the methods to handle big data in text mining. For dimensionality reduction, we should consider the density of data, which has a significant influence on the performance of sentence classification. It requires lots of computations for data of higher dimensions. Eventually, it can cause lots of computational cost and overfitting in the model. Thus, the dimension reduction process is necessary to improve the performance of the model. Diverse methods have been proposed from only lessening the noise of data like misspelling or informal text to including semantic and syntactic information. On top of it, the expression and selection of the text features have impacts on the performance of the classifier for sentence classification, which is one of the fields of Natural Language Processing. The common goal of dimension reduction is to find latent space that is representative of raw data from observation space. Existing methods utilize various algorithms for dimensionality reduction, such as feature extraction and feature selection. In addition to these algorithms, word embeddings, learning low-dimensional vector space representations of words, that can capture semantic and syntactic information from data are also utilized. For improving performance, recent studies have suggested methods that the word dictionary is modified according to the positive and negative score of pre-defined words. The basic idea of this study is that similar words have similar vector representations. Once the feature selection algorithm selects the words that are not important, we thought the words that are similar to the selected words also have no impacts on sentence classification. This study proposes two ways to achieve more accurate classification that conduct selective word elimination under specific regulations and construct word embedding based on Word2Vec embedding. To select words having low importance from the text, we use information gain algorithm to measure the importance and cosine similarity to search for similar words. First, we eliminate words that have comparatively low information gain values from the raw text and form word embedding. Second, we select words additionally that are similar to the words that have a low level of information gain values and make word embedding. In the end, these filtered text and word embedding apply to the deep learning models; Convolutional Neural Network and Attention-Based Bidirectional LSTM. This study uses customer reviews on Kindle in Amazon.com, IMDB, and Yelp as datasets, and classify each data using the deep learning models. The reviews got more than five helpful votes, and the ratio of helpful votes was over 70% classified as helpful reviews. Also, Yelp only shows the number of helpful votes. We extracted 100,000 reviews which got more than five helpful votes using a random sampling method among 750,000 reviews. The minimal preprocessing was executed to each dataset, such as removing numbers and special characters from text data. To evaluate the proposed methods, we compared the performances of Word2Vec and GloVe word embeddings, which used all the words. We showed that one of the proposed methods is better than the embeddings with all the words. By removing unimportant words, we can get better performance. However, if we removed too many words, it showed that the performance was lowered. For future research, it is required to consider diverse ways of preprocessing and the in-depth analysis for the co-occurrence of words to measure similarity values among words. Also, we only applied the proposed method with Word2Vec. Other embedding methods such as GloVe, fastText, ELMo can be applied with the proposed methods, and it is possible to identify the possible combinations between word embedding methods and elimination methods. 텍스트 데이터가 특정 범주에 속하는지 판별하는 문장 분류에서, 문장의 특징을 어떻게 표현하고 어떤 특징을 선택할 것인가는 분류기의 성능에 많은 영향을 미친다. 특징 선택의 목적은 차원을 축소하여도 데이터를 잘설명할 수 있는 방안을 찾아내는 것이다. 다양한 방법이 제시되어 왔으며 Fisher Score나 정보 이득(Information Gain) 알고리즘 등을 통해 특징을 선택 하거나 문맥의 의미와 통사론적 정보를 가지는 Word2Vec 모델로 학습된 단어들을 벡터로 표현하여 차원을 축소하는 방안이 활발하게 연구되었다. 사전에 정의된 단어의 긍정 및 부정 점수에 따라 단어의 임베딩을 수정하는 방법 또한 시도하였다. 본 연구는 문장 분류 문제에 대해 선택적 단어 제거를 수행하고 임베딩을 적용하여 문장 분류 정확도를 향상시키는 방안을 제안한다. 텍스트 데이터에서 정보 이득 값이 낮은 단어들을 제거하고 단어 임베딩을 적용하는방식과, 정보이득 값이 낮은 단어와 코사인 유사도가 높은 주변 단어를 추가로 선택하여 텍스트 데이터에서 제거하고 단어 임베딩을 재구성하는 방식이다. 본 연구에서 제안하는 방안을 수행함에 있어 데이터는 Amazon.com의 ‘Kindle’ 제품에 대한 고객리뷰, IMDB 의 영화리뷰, Yelp의 사용자 리뷰를 사용하였다. Amazon.com의 리뷰 데이터는 유용한 득표수가 5개 이상을 만족하고, 전체 득표 중 유용한 득표의 비율이 70% 이상인 리뷰에 대해 유용한 리뷰라고 판단하였다. Yelp의 경우는 유용한 득표수가 5개 이상인 리뷰 약 75만개 중 10만개를 무작위 추출하였다. 학습에 사용한 딥러닝 모델은 CNN, Attention-Based Bidirectional LSTM을 사용하였고, 단어 임베딩은 Word2Vec과 GloVe를 사용하였다. 단어 제거를 수행하지 않고 Word2Vec 및 GloVe 임베딩을 적용한 경우와 본 연구에서 제안하는 선택적으로 단어 제거를 수행하고 Word2Vec 임베딩을 적용한 경우를 비교하여 통계적 유의성을 검정하였다.

Information Gain-based Initialization for Stable Topological Derivative Segmentation

Choongsang Cho,Nam In Park,Boeun Kim,Younghan Lee,In Hye Yoon 중앙대학교 영상콘텐츠융합연구소 2018 TechArt :Journal of Arts and Imaging Science Vol.5 No.3

This paper describes topological derivatives for image segmentation and restoration. Segmentation performance based on topological derivatives and level set methods heavily depends on initialization information. To achieve stable and accurate segmentation, an efficient initialization scheme is proposed by modeling an image statistically and by analyzing the modeling result based on information theory and classical test theory. Specifically, an image is modeled using a Gaussian mixture model (GMM) for observed data and hidden data; the model information is derived through the maximization of the likelihood distribution and evaluated using the information theory and classical test theory to obtain the weight factors of GMM and class initials. The experimental results demonstrate that the segmentation performance of the proposed method is more stable and accurate than that of the existing algorithms in terms of visual quality and speed.

A Personalized Approach for Recommending Useful Product Reviews Based on Information Gain

최준연,이홍주,박성주 한국인터넷정보학회 2015 KSII Transactions on Internet and Information Syst Vol.9 No.5

Customer product reviews have become great influencers of purchase decision making. To assist potential customers, online stores provide various ways to sort customer reviews. Different methods have been developed to identify and recommend useful reviews to customers, primarily using feedback provided by customers about the helpfulness of reviews. Most of the methods consider the preferences of all users to determine whether reviews are helpful, and all users receive the same recommendations. In this paper, we assessed methods for generating personalized recommendations based on information gain. The information gain approach was extended to consider each individual’s preference together with votes of other users. A total of 172 respondents rated 48 reviews selected from Amazon.com using a 7-point Likert scale. The performance of the devised methods was measured by varying the ratio of training sets and number of recommendations for the data collected. The personalized methods outperformed the existing information gain method, which takes into account the votes from all users. The greatest precision was achieved by the personalized method and a method employing selective use of predictions from the personalized method combined with the existing method based on all users’ reviews. However, the personalized method, which classified helpful reviews based on each user’s threshold value, showed statistically better performance.

불균형 텍스트 데이터의 변수 선택에 있어서의 카이제곱통계량과 정보이득의 특징

Hye In Mun,손원 한국통계학회 2022 응용통계연구 Vol.35 No.4

Since a large text corpus contains hundred-thousand unique words, text data is one of the typical large-dimensional data. Therefore, various feature selection methods have been proposed for dimension reduction. Feature selection methods can improve the prediction accuracy. In addition, with reduced data size, computational efficiency also can be achieved. The chi-square statistic and the information gain are two of the most popular measures for identifying interesting terms from text data. In this paper, we investigate the theoretical properties of the chi-square statistic and the information gain. We show that the two filtering metrics share theoretical properties such as non-negativity and convexity. However, they are different from each other in the sense that the information gain is prone to select more negative features than the chi-square statistic in imbalanced text data. 텍스트 데이터는 일반적으로 많은 단어로 이루어져 있으므로 변수의 수가 매우 많은 고차원 데이터에 해당된다. 이러한 고차원 데이터에서는 계산 효율성과 통계분석의 정확성을 높이기 위해 많은 변수 중 중요한 변수를 선택하기 위한 절차를 거치는 경우가 많다. 텍스트 데이터에서도 많은 단어 중 중요한 단어를 선택하기 위해 여러가지 방법들이 사용되고 있다. 이 연구에서는 단어 선택을 위한 대표적인 필터링 방법인 카이제곱통계량과 정보이득의 공통점과 차이점을 살펴보고 실제 텍스트 데이터에서 이들 성질을 확인해보았다. 카이제곱통계량과 정보이득은 비음성, 볼록성 등의 성질을 공유하지만 불균형 텍스트 데이터에서 카이제곱통계량이 양변수 위주로 단어를 선택하는 반면, 정보이득은 음변수도 상대적으로 많이 선택하는 경향이 있음을 확인하였다.

An Incomplete Information Structure and An Intertemporal General Equilibrium Model of Asset Pricing With Taxes

이일균,Rhee, Il-King Korean Financial Management Association 1991 財務管理硏究 Vol.8 No.2

관측가능 확률과정, 관찰가능변수를 통한 확률과정의 형성과 조세를 중심으로 이 논문은 연속시간의 틀 속에서 재화시장의 수요 및 소비와 생산부문과 자본시장의 수요와 공급을 국민경제에 도입한 일반균형(一般均衡)의 경제분석방법(經濟分析方法)에 의하여 자본자산(資本資産)의 가격(價格)을 결정(決定)하는 일반모형(一般模型)을 제시한다. 이 모형에서는 특히 자본자산의 가격결정에 조세(租稅)가 미치는 영향을 심도있게 분석한다. 이 논문에서는 생산과 소비 그리고 자본자산의 수요와 공급 둥을 결정하는 변수들이 확률과정(確率過程)을 따르는데, 이 변수들을 직접 관찰할 수 있는 경우에 형성되는 자본자산(資本資産)의 가격결정모형(價格決定模型)을 정립한다. 그리고 확률과정의 변수를 직접 관찰할 수 없고 간접적으로 관찰할 수 있을 때에는 간접관찰이 가능한 변수와 확률과정의 변수와의 관계를 정립한 확률과정을 형성하여 자본자산(資本資産)의 가격결정모형(價格決定模型)을 정립한다. 이 모형에는 자산의 가격과 확률적 성질이 모형내에서 결정된다. 이 모형은 증권(證券)의 가격결정(價格決定), 이자율결정(利子率決定), 이자율(利子率)의 기간구조분석(期間構造分析), 이자율(利子率)의 위험구조분석(危險構造分析), 선물가격(先物價格)의 결정(決定) 등 다양하게 이용될 수 있다. This paper develops an intertemporal general equilibrium model of asset pricing with taxes under the noisy and the incomplete information structure and examines theoretically the stochastic behavior of general equilibrium asset prices in a one-good, production, and exchange economy in continuous time markets. The important features of the model are its integration of real and financial markets and the analysis of the effects of differential tax rates between ordinary income and capital gains. The model developed here can provide answers to a wide variety of questions about stochastic structure of asset prices and the effect of tax on them.

한국의 인수합병에 있어서 현금 대금 지급이 주식 대금 지급보다 유리한가?

조성호(Seong Ho Cho) 한국경영학회 2011 經營學硏究 Vol.40 No.5

This paper examines the relationship between the abnormal stock returns upon announcement and the choice of takeover payment methods, say, cash or equity, for the Korean mergers and acquisitions during 1996-2008 in which for some period, say, 1996-2000, liquidity is very scarce and market is uncertain. We test two hypotheses; first, whether equity payments reduce bidders` firm value; second, whether cash payments are more (or less) likely preferred by the target shareholders who are facing high uncertainty. Unlike US experience, we fail to find statistical evidence that equity offers reduce bidders` firm value. Further, while the bidder shareholders gain on average wealth of 1.8% from the M&A transactions, whether they pay in cash or equity, they earn positive 0.1% or 3.5% gain, respectively. These results contradict to the conventional asymmetric information model where market valuation of bidder`s equity plays an important role for the determination of payment alternatives. While examining the wealth gains of target shareholders, we find statistical evidence that cash payments are preferred to equity payments. Further, if they are paid by cash (equity), they earn 6.2% (-5.1%) gain. We conclude that under the liquidity-scarce and uncertain situation, cash is king! Sellers prefer cash rather than taking risk for future higher return. The use of buyer cash would attract more sellers. Buyers prefer equity payments not because of their equity`s overvaluation (conventional model), but because of their cash saving for their lucrative takeovers.

Feature Reduction using a GA-Rough Hybrid Approach on Bio-medical data

Chang Su Lee 제어로봇시스템학회 2011 제어로봇시스템학회 국제학술대회 논문집 Vol.2011 No.10

In this paper, a new approach is proposed for feature reduction using a GA-Rough hybrid approach on Bio-medical data. The given set of bio-medical data is pre-processed with the min-max normalization method. Then the subsequent evaluation on each feature with respect to the output class is carried out utilizing the information gain-based approach using the entropy-based discretization. Features with zero worth on the evaluated set of features are eliminated. The genetic algorithm is applied for performing a search for most relevant features on the set of features remained. These processes continue until there is no further change on the final reduced set of features. The rough set-based approach is applied on this set of features by applying discernibility matrix-based approach in order to obtain the final reduct. The reduced set of features, or a final reduct, is tested for classification using a TS-type rough-fuzzy classifier to show the viability of the proposed feature reduction approach. The results showed that the proposed feature reduction approach effectively achieved to reduce number of features significantly which reduced to 7 out of 120 features along with compatible classification results on the given bio-medical data compared to other approaches.