http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
음식점 리뷰 감성분석을 통한 세부 평가항목별 평점 예측
소진수,신판섭 한국컴퓨터정보학회 2020 韓國컴퓨터情報學會論文誌 Vol.25 No.6
Online reviews we encounter commonly on SNS, although a complex range of assessment information affecting the consumer’s preferences are included, it is general that such information is just provided by simple numbers or star ratings. Based on those review types, it is not easy to get specific information that consumers want and use it to make a decision for purchase. Therefore, in this study, we propose a prediction methodology that can provide ratings broken down by evaluation items by performing sentiment analysis on restaurant reviews written in Korean. To this end, we select ‘food’, ‘price’, ‘service’, and ‘atmosphere’ as the main evaluation items of restaurants, and build a new sentiment dictionary for each evaluation item. It also classifies review sentences by rating item, predicts granular ratings through sentiment analysis, and provides additional information that consumers can use to make decisions. Finally, using MAE and RMSE as evaluation indicators it shows that the rating prediction accuracy of the proposed methodology has been improved than previous studies and presents the use case of proposed methodology. 우리가 SNS상에서 흔하게 접하는 온라인 리뷰에는, 소비자들의 선호도에 영향을 미치는 다양한 평가정보가 복합적으로 포함되어 있지만 이를 매우 간단한 형태의 수치(또는 평점)로 제공하는 것이 일반적이다. 이러한 리뷰에서, 소비자가 원하는 구체적인 정보를 얻고, 이를 구매를 위한 판단에 활용하기란쉽지 않다. 따라서 본 연구에서는 한국어로 작성된 음식점 리뷰를 대상으로, 감성분석을 수행하여 평가항목별로 세분화된 평점을 제공 가능한 예측 방법론을 제안한다. 이를 위해, 음식점의 주요 평가항목으로 ‘음식’, ‘가격’, ‘서비스’, ‘분위기’를 선정하고, 평가항목별 맞춤형 감성사전을 새롭게 구축한다. 또한평가항목별 리뷰 문장을 분류하고 감성분석을 통해 세분화된 평점을 예측하여 소비자가 의사결정에활용 가능한 추가적인 정보를 제공한다. 마지막으로, MAE와 RMSE를 평가지표로 사용하여 기존의연구보다 제안기법의 평점 예측 정확도가 향상되었음을 보이며, 제안 방법론의 활용 사례도 제시한다.
이형주,소진수,김홍근,Lee-Ku Kwac,안원찬 한국물리학회 2019 THE JOURNAL OF THE KOREAN PHYSICAL SOCIETY Vol.75 No.1
The use of an Al$_x$Ga$_{1-x}$As bound Ga$_z$In$_{1-z}$P strain compensation structure for optimum strain in latticed mismatched In$_{0.07}$GaAs/GaAsP$_{0.06}$ multiple quantum wells (MQWs) and its effect on the output power of an infrared light-emitting diode at 940-nm were investigated. A Ga$_{0.53}$InP tensile strain structure, which effectively compensate excessive compressive strain in the In$_{0.07}$GaAs/GaAsP$_{0.06}$ MQWs, was inserted between a quantum well and a quantum barrier. The Al$_{0.2}$GaAs material was used as both a growth buffer and a balancing barrier for In$_{0.07}$GaAs/Al$_x$Ga$_{1-x}$As-bound Ga$_{0.53}$InP/GaAsP$_{0.06}$ MQWs. From photoluminescence (PL) measurements and X-ray diffraction (XRD) rocking curves, we verified that the Ga$_{0.53}$InP tensile strain barrier could effectively compensate the compressive strain of the In$_{0.07}$GaAs/GaAsP$_{0.06}$ MQWs. In addition, a further increase in the PL intensity from the In$_{0.07}$GaAs/Al$_y$Ga$_{1-y}$As-bound Ga$_{0.53}$InP/GaAsP$_{0.06}$ MQWs was found after having adjusted the Al$_{0.2}$GaAs strain tuning barrier. This result was significantly supported by the stable balance of the energy bandgap structure in the developed MQWs. From fabricated IR-LEDs chips, the LED with an In$_{0.07}$GaAs/GaAsP$_{0.06}$ MQW employing the Al$_{0.2}$GaAs-bound Ga$_{0.53}$InP strain compensation structure displayed a 48\% higher light output power as compared with a conventional LED. These results suggest that the use of an Al$_{0.2}$GaAs-bound Ga$_{0.53}$InP strain compensation structure effectively improved both the unbalanced strain and the unbalanced energy bandgap of lattice-mismatched In$_{0.07}$GaAs/GaAsP$_{0.06}$ MQWs for 940-nm IR-LEDs.
이형주,박광훈,소진수,김재훈,김홍근,곽이규 한국물리학회 2021 Current Applied Physics Vol.22 No.-
In this paper, we report the synthesis and transmittance of a titanium–indium–tin oxide (TITO) film, fabricated through a low-temperature process. The TITO film was fabricated by incorporating a 2-nm-thick titanium barrier at the bottom of an ITO film. The transmittance characteristics of the TITO film were examined for light-emitting diodes (LEDs) of various wavelengths at different post-annealing temperatures. A saturated high transmittance was observed at a temperature of 550 ◦C, which is relatively low when compared to that in the case of a conventional ITO film. Photoluminescence studies demonstrated that a 450-nm-thick TITO film, fabricated at 550 ◦C, was highly effective in improving the performance of the LED, when compared to conventional ITO films. The X-ray diffraction peaks, scanning electron microscopy images, and transmittance electron microscopy images confirmed that titanium atoms could improve the crystallization of ITO. It was found that non-crystallization in ITO was effectively activated by the titanium barrier. Furthermore, the optical bandgap (3.77 eV for the conventional ITO film) was improved to 3.92 eV in the TITO film. An infrared LED fabricated with a TITO film displayed 70% higher light output power than that with a conventional ITO film. These results suggest that using a titanium barrier is essential to effectively improve inactive nucleation sites in ITO films grown at low temperatures.
텍스트 분석을 이용한 청원데이터의 주제 및 감성에 관한 연구
서혜선(Hyesun Suh),소진수(Jinsoo So) 한국자료분석학회 2020 Journal of the Korean Data Analysis Society Vol.22 No.3
본 연구는 국민들의 관심사와 참여가 많은 청와대 국민청원 게시글에 대한 텍스트 분석과 토픽 모델링을 실시하는데 있다. 17개로 분류되어져 있는 국민청원데이터는 개설된 이래 60여만개의 게시글이 올라와 빅데이터 수준으로 축적되어 있는 상황이다. 이러한 청원 게시글에 대한 최초의 17개 분류가 적정한지를 텍스트 분석 기반으로 재분류해 보고자 한다. 이를 통해 국민청원 게시글에 대한 새로운 토픽(분류, 주제)들을 제안하고 단어들의 긍정어, 부정어 등을 고려한 감성 분석을 통해 각 토픽별 감성의 수준을 제시하고자 한다. 또한 선정된 토픽별로 청원글에 대한 국민참여수가 달라지는지, 그리고 각 청원글에 포함된 단어들의 감성 수준에 따라 국민참여수가 달라지는가를 연구하기 위해 일반화선형모형을 적용한 분석을 실시한다. 분석 결과 토픽과 감성에 따라 청원 게시글에 대한 국민동참의 참여도가 통계적 유의성하에서 달라짐을 확인하였다. 또한 청원 게시글에 참여를 올리는 데는 부정어의 사용보다 긍정어의 사용이 더 효과적이라는 사실을 확인할 수 있었다. The purpose of this study is to conduct text analysis and topic modeling on the posts of the National Petition of the Blue House, where there are many interests and participation of the people. The national petition data, which has been classified into 17, has accumulated over 600,000 posts since it was opened and accumulated at the level of big data. We would like to reclassify whether the first 17 classifications of these petitions are appropriate based on text analysis. Through this, we propose new topics (classifications) for the postings of the National Petition, and suggest sentiment levels for each topic through sentiment analysis considering the positive and negative words. In addition, analysis using generalized linear model (GLM) is conducted to study whether there is a difference in the participation number of the national participation posts according to the topics and the sentiments of words. As a result of the analysis, it was confirmed that the participation rate of citizens in the petition posts was different under statistical significance according to the topics and sentiments. Also, it was confirmed that the use of positive words was more effective than the use of negative words to increase participation in the petition post.