• Journal of Intelligence and Information Systems
• /
• v.26 no.4
• /
• pp.1-25
• /
• 2020

Sentiment Analysis (SA) is a Natural Language Processing (NLP) task that analyzes the sentiments consumers or the public feel about an arbitrary object from written texts. Furthermore, Aspect-Based Sentiment Analysis (ABSA) is a fine-grained analysis of the sentiments towards each aspect of an object. Since having a more practical value in terms of business, ABSA is drawing attention from both academic and industrial organizations. When there is a review that says "The restaurant is expensive but the food is really fantastic", for example, the general SA evaluates the overall sentiment towards the 'restaurant' as 'positive', while ABSA identifies the restaurant's aspect 'price' as 'negative' and 'food' aspect as 'positive'. Thus, ABSA enables a more specific and effective marketing strategy. In order to perform ABSA, it is necessary to identify what are the aspect terms or aspect categories included in the text, and judge the sentiments towards them. Accordingly, there exist four main areas in ABSA; aspect term extraction, aspect category detection, Aspect Term Sentiment Classification (ATSC), and Aspect Category Sentiment Classification (ACSC). It is usually conducted by extracting aspect terms and then performing ATSC to analyze sentiments for the given aspect terms, or by extracting aspect categories and then performing ACSC to analyze sentiments for the given aspect category. Here, an aspect category is expressed in one or more aspect terms, or indirectly inferred by other words. In the preceding example sentence, 'price' and 'food' are both aspect categories, and the aspect category 'food' is expressed by the aspect term 'food' included in the review. If the review sentence includes 'pasta', 'steak', or 'grilled chicken special', these can all be aspect terms for the aspect category 'food'. As such, an aspect category referred to by one or more specific aspect terms is called an explicit aspect. On the other hand, the aspect category like 'price', which does not have any specific aspect terms but can be indirectly guessed with an emotional word 'expensive,' is called an implicit aspect. So far, the 'aspect category' has been used to avoid confusion about 'aspect term'. From now on, we will consider 'aspect category' and 'aspect' as the same concept and use the word 'aspect' more for convenience. And one thing to note is that ATSC analyzes the sentiment towards given aspect terms, so it deals only with explicit aspects, and ACSC treats not only explicit aspects but also implicit aspects. This study seeks to find answers to the following issues ignored in the previous studies when applying the BERT pre-trained language model to ACSC and derives superior ACSC models. First, is it more effective to reflect the output vector of tokens for aspect categories than to use only the final output vector of [CLS] token as a classification vector? Second, is there any performance difference between QA (Question Answering) and NLI (Natural Language Inference) types in the sentence-pair configuration of input data? Third, is there any performance difference according to the order of sentence including aspect category in the QA or NLI type sentence-pair configuration of input data? To achieve these research objectives, we implemented 12 ACSC models and conducted experiments on 4 English benchmark datasets. As a result, ACSC models that provide performance beyond the existing studies without expanding the training dataset were derived. In addition, it was found that it is more effective to reflect the output vector of the aspect category token than to use only the output vector for the [CLS] token as a classification vector. It was also found that QA type input generally provides better performance than NLI, and the order of the sentence with the aspect category in QA type is irrelevant with performance. There may be some differences depending on the characteristics of the dataset, but when using NLI type sentence-pair input, placing the sentence containing the aspect category second seems to provide better performance. The new methodology for designing the ACSC model used in this study could be similarly applied to other studies such as ATSC.

• Journal of the Daesoon Academy of Sciences
• /
• v.33
• /
• pp.31-62
• /
• 2019

The purpose of this article is to study on the Daesoon Cosmology of the Correlative Relation between Mugeuk and Taegeuk. Daesoon cosmology is a cosmology based on the juxtaposition between the Gucheon Sangje and the world. In this article, I would like to say that this theory in Daesoon Thought was developed in three stages: the phase of the Mugeuk Transcendence of Gucheon Sangje, the stage of the Taegeuk Immanence, and the phase of the Grand Opening of the Later World between Mugeuk and Taegeuk as a correlative gentle reign. First of all, the phase of the Mugeuk Transcendence of Gucheon Sangje has been revealed as a yin-yang relationship. The stage of the Taegeuk Immanence represents the togetherness of harmony and co-prosperity between yin and yang, and the phase of the Grand Opening of the Later World between Mukeuk and Taegeuk refers to the unshakable accomplishment of its character and energy. It will be said that this is due to the practical mechanism in the correct balance of yin-yang making a four stage cycle of birth, growth, harvest, and storage. In addition, the Daesoon stage of the settlement of yin and yang is revealed as a change in the growth of all things and the formation of the inner circle. The mental growth reveals the characteristics of everything in the world, each trying to shine at the height of their own respective life as they grow up energetically. The dominant culture of cerebral communion renders a soft and elegant mood and combines yin and yang to elevate the heavenly and earthly period through transcendental change into sympathetic understanding. The stage of the Grand Opening of the Later World between Mugeuk and Taegeuk is one of the earliest days of the lunar month and also the inner circle of Taegeuk. It is in line with Ken Wilbur's integrated model as a step to the true degrees to develop into a world with brightened degrees. It is a beautiful and peaceful scene where celestial maidens play music, the firewood burns, and the scholars command thunder and lightning playfully. Human beings achieve a state of happiness as a free beings who lives as gods upon the earth. This is the world of theGrand Opening of the Later World between Mugeuk and Taegeuk. Daesoon Thought was succeeded by Dojeon in 1958, when Dojeon emerged as the successor in the lineage of religious orthodoxy and was assigned the task of handling Dao in its entirety. In addition, Daesoon is a circle and represents freedom and commonly shared happiness among the populous. Cosmology in the Daesoon Thought will enable us to understand deep dimensions and the identity of members as individuals within an inner circle of correlation between transcendence and immanence. This present study tries to analyze the public effects philologically and also the mutual correlation by utilizing the truthfulness of literature and rational interpretation. The outlook for the future in Daesoon Thought also leads to the one-way communication of Daesoon as a circle.

• Korean small business review
• /
• v.32 no.2
• /
• pp.209-223
• /
• 2010

Korea's business is composed of a few large-sized enterprises (which can be abbreviated as LSE) and a majority of small-sized enterprises (SSE). Although there has been a growing recognition of the need for the development of medium-sized enterprises (MSE) which can serve as a link between SSE and LSE, as yet there has not yet been a consensus on the definition, characteristics and the function of the MSE in Korea. Nowadays, the world is being globalized, and Japan and China are in competition to ne a great economic power. While East Asia is experiencing rapid changes, promoting MSE which can secure flexibility and efficiency through covering up the limitation of LSE and SSE is needed in order to respond the global market which is being specialized. The features of MSE in Japan can be listed as follows. First, the MSE in Japan is developing the company through getting into niche markets which are hard for major companies to enter rather than developing markets in order to compete against major companies directly. While MSEs are endeavoring to build the business firmly in the domestic market, they can possess special and competitive technical skills through trials and errors; so that they can get a chance develop their business through independent business system rather than putting their effort to compete against major companies. Second, from the MSEs with competitive edge in the market, there are many contributions to the national exportation. Those MSEs produce in domestic and maintain the quality of high price products which need cutting-edge technology, while they relocate the low and middle priced goods to the country where manufacturing costs are low, so that they can maintain the price competitiveness. Third, the industrial structure in Japan is formed from dual structure between major companies and small sized companies. In other words, in Japan's industrial structure which are composed of subcontract structure, this dual structure has taken a major role of small sized companies' growth and manufacturing businesses' international competitive power. Forth, MSE in Japan adopt a strategy of putting their value on qualitative scale growth rather than quantitative scale growth. In this paper, the case of Japanese MSE is analyzed. Along with its long history of Industrialization, Japan has a corporate environment where the SSEs can develop as a MSE and later a LSE through a full-support system. Among its SSEs, there are a number of world class corporations equipped with a large domestic market, win-win cooperation with the LSEs and an independent technology development. It can also be observed that these SSEs develop into MSEs with sustainable growth potentials. This study will focus on the condition under which the MSEs of Japan have been developed, and how they have survived the competition between SSEs and LSEs. Through this study, this paper attempts to offer solutions to Korea's polarization between the SSE and LSE, while providing the basis for SSEs revitalization. In general, if both extremities phenomenon deepen between LSE and SSE, there are possible fears of occurring disutility in national economy by the monopolization of LSE. For that reason, enterprise group, which can make SSE or MSE compete LSE in some area and ease the monopoly and oligopoly problem, is needed. This awareness has been shared for ages long. Nevertheless, there is no legal definition for MSE in Japan, and there is no definition about the enterprise size or unified view of MSE between scholars, but it is defined differently by each of academical person or research institution and study meeting. For that reason, this paper will organize the definition of MSE in Japan, and then will propose the characteristics of the background which has made MSE secure competitiveness and sustainable growth in global market. This study focus on that because through this process, the positive change to the awareness of MSE can be proposed in Korea and to seek the policy direction for building institutional framework which can make SSE become MES. Through this way, the fundamentals for SSE to become MSE can be managed and some appropriate suggestions which will be able to make MSE enter the global market in the future can also be proposed. Due to these facts, this study is very important and well timed task. In a sense of this way, this study will examine the definition and role of MSE in Japan. after this examination, this study will deal with the status, special feature, and promotion policy for MSE. Through this analysis of MSE in Japan, the foundation which be able to set the desirable role model for MSE in Korea can be proposed. Also, the political implication which is needed to push ahead to contribute to creating employment and economic growth through sustainable growth of MSEs in economic system of Korea can be offered through this study. It has been found that Japan's MSE functions as an indispensable link among various industrial structures by holding a significant position in employment rate, production and value added. Although the MSEs took up less than 1% of the entire number of businesses with 2700 manufacturing firms and 7000 non-manufacturing firms, its employment ratios are about 15%, while taking about 25% of the manufacturing industry's exports. In industries such as machinery and electronics which is considered Japan's major industry, the MSEs showed a higher than average ratio of manufacturing exports and employment rate. It can be analyzed that behind Japan's advantageous industries, close and deeply knit MSEs exist. Although there are no clearly stated policies geared towards the MSEs by the Japanese government, various political measures exist such as the R&D Project and the inducement of cooperation between enterprises which gives room for MSEs to participate in the SSE policies. In relation to these findings, the following practical measures can be considered in order to revitalize Korea's MSEs: First, there is a need for a legal definition of MSE and the incentives to provide legal support for its growth. Second, if a law to support the MSEs is established, it could provide a powerful inducement for the SSE to grow as a MSE, rather than stay as a SSE. Third, there is a need for a strategy of MSEs to establish a stable base in the domestic market and then advance to the global market with the accumulated trial and error and competitiveness. Fourth, the SSE themselves need the spirit of entrepreneurship in order to make the leap to a MSE. Because if nothing is to be changed about the system on the firms that grew, and the parts of the past custom was left to be managed alone, confusion and absence of management can take place. No matter how much tax favors the government will give and no matter how much incentive there could be through the policies, there are limits for industries to higher the ability to propagate. And because of that it is a period where industries need their own innovative skills to reform their firms.

• Journal of Intelligence and Information Systems
• /
• v.24 no.1
• /
• pp.205-225
• /
• 2018

Convolutional Neural Network (ConvNet) is one class of the powerful Deep Neural Network that can analyze and learn hierarchies of visual features. Originally, first neural network (Neocognitron) was introduced in the 80s. At that time, the neural network was not broadly used in both industry and academic field by cause of large-scale dataset shortage and low computational power. However, after a few decades later in 2012, Krizhevsky made a breakthrough on ILSVRC-12 visual recognition competition using Convolutional Neural Network. That breakthrough revived people interest in the neural network. The success of Convolutional Neural Network is achieved with two main factors. First of them is the emergence of advanced hardware (GPUs) for sufficient parallel computation. Second is the availability of large-scale datasets such as ImageNet (ILSVRC) dataset for training. Unfortunately, many new domains are bottlenecked by these factors. For most domains, it is difficult and requires lots of effort to gather large-scale dataset to train a ConvNet. Moreover, even if we have a large-scale dataset, training ConvNet from scratch is required expensive resource and time-consuming. These two obstacles can be solved by using transfer learning. Transfer learning is a method for transferring the knowledge from a source domain to new domain. There are two major Transfer learning cases. First one is ConvNet as fixed feature extractor, and the second one is Fine-tune the ConvNet on a new dataset. In the first case, using pre-trained ConvNet (such as on ImageNet) to compute feed-forward activations of the image into the ConvNet and extract activation features from specific layers. In the second case, replacing and retraining the ConvNet classifier on the new dataset, then fine-tune the weights of the pre-trained network with the backpropagation. In this paper, we focus on using multiple ConvNet layers as a fixed feature extractor only. However, applying features with high dimensional complexity that is directly extracted from multiple ConvNet layers is still a challenging problem. We observe that features extracted from multiple ConvNet layers address the different characteristics of the image which means better representation could be obtained by finding the optimal combination of multiple ConvNet layers. Based on that observation, we propose to employ multiple ConvNet layer representations for transfer learning instead of a single ConvNet layer representation. Overall, our primary pipeline has three steps. Firstly, images from target task are given as input to ConvNet, then that image will be feed-forwarded into pre-trained AlexNet, and the activation features from three fully connected convolutional layers are extracted. Secondly, activation features of three ConvNet layers are concatenated to obtain multiple ConvNet layers representation because it will gain more information about an image. When three fully connected layer features concatenated, the occurring image representation would have 9192 (4096+4096+1000) dimension features. However, features extracted from multiple ConvNet layers are redundant and noisy since they are extracted from the same ConvNet. Thus, a third step, we will use Principal Component Analysis (PCA) to select salient features before the training phase. When salient features are obtained, the classifier can classify image more accurately, and the performance of transfer learning can be improved. To evaluate proposed method, experiments are conducted in three standard datasets (Caltech-256, VOC07, and SUN397) to compare multiple ConvNet layer representations against single ConvNet layer representation by using PCA for feature selection and dimension reduction. Our experiments demonstrated the importance of feature selection for multiple ConvNet layer representation. Moreover, our proposed approach achieved 75.6% accuracy compared to 73.9% accuracy achieved by FC7 layer on the Caltech-256 dataset, 73.1% accuracy compared to 69.2% accuracy achieved by FC8 layer on the VOC07 dataset, 52.2% accuracy compared to 48.7% accuracy achieved by FC7 layer on the SUN397 dataset. We also showed that our proposed approach achieved superior performance, 2.8%, 2.1% and 3.1% accuracy improvement on Caltech-256, VOC07, and SUN397 dataset respectively compare to existing work.