• Journal of Intelligence and Information Systems
• /
• v.24 no.2
• /
• pp.125-148
• /
• 2018

Recently, as the demand for big data analysis increases, cases of analyzing unstructured data and using the results are also increasing. Among the various types of unstructured data, text is used as a means of communicating information in almost all fields. In addition, many analysts are interested in the amount of data is very large and relatively easy to collect compared to other unstructured and structured data. Among the various text analysis applications, document classification which classifies documents into predetermined categories, topic modeling which extracts major topics from a large number of documents, sentimental analysis or opinion mining that identifies emotions or opinions contained in texts, and Text Summarization which summarize the main contents from one document or several documents have been actively studied. Especially, the text summarization technique is actively applied in the business through the news summary service, the privacy policy summary service, ect. In addition, much research has been done in academia in accordance with the extraction approach which provides the main elements of the document selectively and the abstraction approach which extracts the elements of the document and composes new sentences by combining them. However, the technique of evaluating the quality of automatically summarized documents has not made much progress compared to the technique of automatic text summarization. Most of existing studies dealing with the quality evaluation of summarization were carried out manual summarization of document, using them as reference documents, and measuring the similarity between the automatic summary and reference document. Specifically, automatic summarization is performed through various techniques from full text, and comparison with reference document, which is an ideal summary document, is performed for measuring the quality of automatic summarization. Reference documents are provided in two major ways, the most common way is manual summarization, in which a person creates an ideal summary by hand. Since this method requires human intervention in the process of preparing the summary, it takes a lot of time and cost to write the summary, and there is a limitation that the evaluation result may be different depending on the subject of the summarizer. Therefore, in order to overcome these limitations, attempts have been made to measure the quality of summary documents without human intervention. On the other hand, as a representative attempt to overcome these limitations, a method has been recently devised to reduce the size of the full text and to measure the similarity of the reduced full text and the automatic summary. In this method, the more frequent term in the full text appears in the summary, the better the quality of the summary. However, since summarization essentially means minimizing a lot of content while minimizing content omissions, it is unreasonable to say that a "good summary" based on only frequency always means a "good summary" in its essential meaning. In order to overcome the limitations of this previous study of summarization evaluation, this study proposes an automatic quality evaluation for text summarization method based on the essential meaning of summarization. Specifically, the concept of succinctness is defined as an element indicating how few duplicated contents among the sentences of the summary, and completeness is defined as an element that indicating how few of the contents are not included in the summary. In this paper, we propose a method for automatic quality evaluation of text summarization based on the concepts of succinctness and completeness. In order to evaluate the practical applicability of the proposed methodology, 29,671 sentences were extracted from TripAdvisor 's hotel reviews, summarized the reviews by each hotel and presented the results of the experiments conducted on evaluation of the quality of summaries in accordance to the proposed methodology. It also provides a way to integrate the completeness and succinctness in the trade-off relationship into the F-Score, and propose a method to perform the optimal summarization by changing the threshold of the sentence similarity.

• Journal of Intelligence and Information Systems
• /
• v.19 no.4
• /
• pp.123-132
• /
• 2013

As the smartphones are equipped with various sensors such as the accelerometer, GPS, gravity sensor, gyros, ambient light sensor, proximity sensor, and so on, there have been many research works on making use of these sensors to create valuable applications. Human activity recognition is one such application that is motivated by various welfare applications such as the support for the elderly, measurement of calorie consumption, analysis of lifestyles, analysis of exercise patterns, and so on. One of the challenges faced when using the smartphone sensors for activity recognition is that the number of sensors used should be minimized to save the battery power. When the number of sensors used are restricted, it is difficult to realize a highly accurate activity recognizer or a classifier because it is hard to distinguish between subtly different activities relying on only limited information. The difficulty gets especially severe when the number of different activity classes to be distinguished is very large. In this paper, we show that a fairly accurate classifier can be built that can distinguish ten different activities by using only a single sensor data, i.e., the smartphone accelerometer data. The approach that we take to dealing with this ten-class problem is to use the ensemble of nested dichotomy (END) method that transforms a multi-class problem into multiple two-class problems. END builds a committee of binary classifiers in a nested fashion using a binary tree. At the root of the binary tree, the set of all the classes are split into two subsets of classes by using a binary classifier. At a child node of the tree, a subset of classes is again split into two smaller subsets by using another binary classifier. Continuing in this way, we can obtain a binary tree where each leaf node contains a single class. This binary tree can be viewed as a nested dichotomy that can make multi-class predictions. Depending on how a set of classes are split into two subsets at each node, the final tree that we obtain can be different. Since there can be some classes that are correlated, a particular tree may perform better than the others. However, we can hardly identify the best tree without deep domain knowledge. The END method copes with this problem by building multiple dichotomy trees randomly during learning, and then combining the predictions made by each tree during classification. The END method is generally known to perform well even when the base learner is unable to model complex decision boundaries As the base classifier at each node of the dichotomy, we have used another ensemble classifier called the random forest. A random forest is built by repeatedly generating a decision tree each time with a different random subset of features using a bootstrap sample. By combining bagging with random feature subset selection, a random forest enjoys the advantage of having more diverse ensemble members than a simple bagging. As an overall result, our ensemble of nested dichotomy can actually be seen as a committee of committees of decision trees that can deal with a multi-class problem with high accuracy. The ten classes of activities that we distinguish in this paper are 'Sitting', 'Standing', 'Walking', 'Running', 'Walking Uphill', 'Walking Downhill', 'Running Uphill', 'Running Downhill', 'Falling', and 'Hobbling'. The features used for classifying these activities include not only the magnitude of acceleration vector at each time point but also the maximum, the minimum, and the standard deviation of vector magnitude within a time window of the last 2 seconds, etc. For experiments to compare the performance of END with those of other methods, the accelerometer data has been collected at every 0.1 second for 2 minutes for each activity from 5 volunteers. Among these 5,900 ($=5{\times}(60{\times}2-2)/0.1$) data collected for each activity (the data for the first 2 seconds are trashed because they do not have time window data), 4,700 have been used for training and the rest for testing. Although 'Walking Uphill' is often confused with some other similar activities, END has been found to classify all of the ten activities with a fairly high accuracy of 98.4%. On the other hand, the accuracies achieved by a decision tree, a k-nearest neighbor, and a one-versus-rest support vector machine have been observed as 97.6%, 96.5%, and 97.6%, respectively.

• Korean Journal of Heritage: History & Science
• /
• v.52 no.2
• /
• pp.142-171
• /
• 2019

The dragon is an imaginary animal that appears in the legends and myths of the Orient and the West. While dragons have mostly been portrayed as aggressive and as bad omens in the West, in the Orient, as they symbolize the emperor or have an auspicious meaning, dragons signify a positive meaning. In addition, as the dragon symbolizes the emperor and its type has been diversified considering it as a divine object that controls water, people have tried to express it as a figure. The records related to dragons in the Goryeo dynasty appeared with diverse topics in 'History of Goryeo' and are generally contents related to founding myths, rituals for rain, and Shinii (神異), etc. The founding myth emphasizes the legality of the Goryeo dynasty through the dragon, and this influenced the formation of the dragon's descendants. In addition, the ability to control water, which is a characteristic of the dragon, was symbolized as an earth dragon related to the rainmaking ritual, i.e., wishing for rain during times of drought. Since the dragon was the symbol of the royal family, the use of the dragon by common people was strictly restricted. Furthermore, the association of a bronze dragon mirror with the royal family is hard to be excluded. The type and quantity of bronze double dragon mirrors discovered to have existed during the Goryeo dynasty is great, and the production and the distribution of bronze mirrors with double dragons seem to have been more active compared to other bronze mirrors, as bronze mirrors with double dragons produced during Goryeo and bronze mirrors originating in China were mixed. Therefore, in this article, the characteristics of diverse bronze mirrors from the 10th century to the 14th century in China were examined. It seems that the master craftsmen who produced bronze mirrors with double dragons during the Goryeo dynasty were influenced by Chinese composition patterns when making the mirrors. Because there were many cases where a bronze mirror's country of origin could not easily be determined, in order to identify the differences between bronze double dragon mirrors produced during the Goryeo dynasty and bronze mirrors produced in China, meticulous analysis was required. Thus, to ascertain that Goryeo mirrors were not imitations of bronze mirrors with double dragons originating in China but produced independently, the mirrors were examined using the bronze double dragon mirror type classification system existing in our country. Bronze mirrors with double dragons are classified into three types: Type I, which has the style of the Yao dynasty, includes the greatest proportion; however, despite there being only a small quantity for comparison, Types II and III were selected for the analysis of the bronze mirrors with double dragons made in Goryeo because they have unique composition patterns. As mentioned above, distinguishing bronze mirrors made during Goryeo from bronze mirrors made in China is challenging because Goryeo bronze mirrors were made under the influence of China. Among them, since the manufacturing place of the bronze mirrors with double dragons found at the nine-story stone pagoda in Woljeongsa Temple in Pyeongchang is questionable and the composition pattern of the bronze mirror is hard to find on bronze mirrors with double dragons made in China, the manufacturing place of those bronze mirrors were examined. These bronze mirrors with double dragons were considered as bronze mirrors with double dragons made during the Goryeo dynasty adopting the Yao dynasty style composition pattern as aspects of the composition pattern belonged to Type I, and the detailed combination of patterns is hard to find in mirrors produced in China.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.21-41
• /
• 2019

Thanks to the rapid development of information technologies, the data available on Internet have grown rapidly. In this era of big data, many studies have attempted to offer insights and express the effects of data analysis. In the tourism and hospitality industry, many firms and studies in the era of big data have paid attention to online reviews on social media because of their large influence over customers. As tourism is an information-intensive industry, the effect of these information networks on social media platforms is more remarkable compared to any other types of media. However, there are some limitations to the improvements in service quality that can be made based on opinions on social media platforms. Users on social media platforms represent their opinions as text, images, and so on. Raw data sets from these reviews are unstructured. Moreover, these data sets are too big to extract new information and hidden knowledge by human competences. To use them for business intelligence and analytics applications, proper big data techniques like Natural Language Processing and data mining techniques are needed. This study suggests an analytical approach to directly yield insights from these reviews to improve the service quality of hotels. Our proposed approach consists of topic mining to extract topics contained in the reviews and the decision tree modeling to explain the relationship between topics and ratings. Topic mining refers to a method for finding a group of words from a collection of documents that represents a document. Among several topic mining methods, we adopted the Latent Dirichlet Allocation algorithm, which is considered as the most universal algorithm. However, LDA is not enough to find insights that can improve service quality because it cannot find the relationship between topics and ratings. To overcome this limitation, we also use the Classification and Regression Tree method, which is a kind of decision tree technique. Through the CART method, we can find what topics are related to positive or negative ratings of a hotel and visualize the results. Therefore, this study aims to investigate the representation of an analytical approach for the improvement of hotel service quality from unstructured review data sets. Through experiments for four hotels in Hong Kong, we can find the strengths and weaknesses of services for each hotel and suggest improvements to aid in customer satisfaction. Especially from positive reviews, we find what these hotels should maintain for service quality. For example, compared with the other hotels, a hotel has a good location and room condition which are extracted from positive reviews for it. In contrast, we also find what they should modify in their services from negative reviews. For example, a hotel should improve room condition related to soundproof. These results mean that our approach is useful in finding some insights for the service quality of hotels. That is, from the enormous size of review data, our approach can provide practical suggestions for hotel managers to improve their service quality. In the past, studies for improving service quality relied on surveys or interviews of customers. However, these methods are often costly and time consuming and the results may be biased by biased sampling or untrustworthy answers. The proposed approach directly obtains honest feedback from customers' online reviews and draws some insights through a type of big data analysis. So it will be a more useful tool to overcome the limitations of surveys or interviews. Moreover, our approach easily obtains the service quality information of other hotels or services in the tourism industry because it needs only open online reviews and ratings as input data. Furthermore, the performance of our approach will be better if other structured and unstructured data sources are added.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.179-196
• /
• 2019

Recently, the "Smart Consumer" has been emerging. He or she is increasingly inclined to search for and purchase products by taking into account personal judgment or expert reviews rather than by relying on information delivered through manufacturers' advertising. This is especially true when purchasing cosmetics. Because cosmetics act directly on the skin, consumers respond seriously to dangerous chemical elements they contain or to skin problems they may cause. Above all, cosmetics should fit well with the purchaser's skin type. In addition, changes in global cosmetics consumer trends make it necessary to study this field. The desire to find one's own individualized cosmetics is being revealed to consumers around the world and is known as "Finding the Holy Grail." Many consumers show a deep interest in customized cosmetics with the cultural boom known as "K-Beauty" (an aspect of "Han-Ryu"), the growth of personal grooming, and the emergence of "self-culture" that includes "self-beauty" and "self-interior." These trends have led to the explosive popularity of cosmetics made in Korea in the Chinese and Southeast Asian markets. In order to meet the customized cosmetics needs of consumers, cosmetics manufacturers and related companies are responding by concentrating on delivering premium services through the convergence of ICT(Information, Communication and Technology). Despite the evolution of companies' responses regarding market trends toward customized cosmetics, there is no "Intelligent Data Platform" that deals holistically with consumers' skin condition experience and thus attaches emotions to products and services. To find the Holy Grail of customized cosmetics, it is important to acquire and analyze consumer data on what they want in order to address their experiences and emotions. The emotions consumers are addressing when purchasing cosmetics varies by their age, sex, skin type, and specific skin issues and influences what price is considered reasonable. Therefore, it is necessary to classify emotions regarding cosmetics by individual consumer. Because of its importance, consumer emotion analysis has been used for both services and products. Given the trends identified above, we judge that consumer emotion analysis can be used in our study. Therefore, we collected and indexed data on consumers' emotions regarding their cosmetics experiences focusing on consumers' language. We crawled the cosmetics emotion data from SNS (blog and Twitter) according to sales ranking ($1^{st}$ to $99^{th}$), focusing on the ample/serum category. A total of 357 emotional adjectives were collected, and we combined and abstracted similar or duplicate emotional adjectives. We conducted a "Consumer Sentiment Journey" workshop to build a "Consumer Sentiment Dictionary," and this resulted in a total of 76 emotional adjectives regarding cosmetics consumer experience. Using these 76 emotional adjectives, we performed clustering with the Self-Organizing Map (SOM) method. As a result of the analysis, we derived eight final clusters of cosmetics consumer sentiments. Using the vector values of each node for each cluster, the characteristics of each cluster were derived based on the top ten most frequently appearing consumer sentiments. Different characteristics were found in consumer sentiments in each cluster. We also developed a cosmetics experience pattern map. The study results confirmed that recommendation and classification systems that consider consumer emotions and sentiments are needed because each consumer differs in what he or she pursues and prefers. Furthermore, this study reaffirms that the application of emotion and sentiment analysis can be extended to various fields other than cosmetics, and it implies that consumer insights can be derived using these methods. They can be used not only to build a specialized sentiment dictionary using scientific processes and "Design Thinking Methodology," but we also expect that these methods can help us to understand consumers' psychological reactions and cognitive behaviors. If this study is further developed, we believe that it will be able to provide solutions based on consumer experience, and therefore that it can be developed as an aspect of marketing intelligence.

• Journal of Food Hygiene and Safety
• /
• v.34 no.3
• /
• pp.251-262
• /
• 2019

In this study, we analyzed Korean dietary habits with food intake data from the Korea National Health and Nutrition Examination Survey (KNHANES) and the Korea Centers for Disease Control and Prevention and we proposed a set of management guidelines for future Korean dietary habits. A total of 839 food items (1,419 foods) were analyzed according to the food catagories in "Food Code", which is the representative food classification system in Korea. The average total daily food intake was 1,585.77 g/day, with raw and processed foods accounting for 858.96 g/day and 726.81 g/day, respectively. Cereal grains contributed to the highest proportion of the food intake. Over 90% of subjects consumed cereal grains (99.09%) and root and tuber vegetables (95.80%) among the top 15 consumed food groups. According to the analysis by item, rice, Korean cabbage kimchi, apple, radish, egg, chili pepper, onion, wheat, soybean curds, potato, cucumber and pork were major (at least 1% of the average daily intake, 158.6 g/day) and frequently (eaten by more than 25% of subjects, 5,168 persons) consumed food items, and Korean spices were at the top of this list. In the case of kimchi, the proportion of intake of Korean cabbage kimchi (64.89 g/day) was the highest. In the case of alcoholic beverages, intake was highest by order of beer (63.53 g/day), soju (39.11 g/day) and makgeolli (19.70 g/day), and intake frequency was high in order of soju (11.3%), beer (7.2%), and sake (6.6%). Analysis results by seasonal intake trends showed that cereal grains have steadily decreased and beverages have slightly risen. In the case of alcoholic beverage consumption frequency, some kinds of makgeolli, wine, sake, and black raspberry wine have decreased gradually year by year. The consumption trend for kimchi has been gradually decreasing as well.

• Journal of Intelligence and Information Systems
• /
• v.26 no.4
• /
• pp.111-126
• /
• 2020

Most industries have recently become aware of the importance of customer lifetime value as they are exposed to a competitive environment. As a result, preventing customers from churn is becoming a more important business issue than securing new customers. This is because maintaining churn customers is far more economical than securing new customers, and in fact, the acquisition cost of new customers is known to be five to six times higher than the maintenance cost of churn customers. Also, Companies that effectively prevent customer churn and improve customer retention rates are known to have a positive effect on not only increasing the company's profitability but also improving its brand image by improving customer satisfaction. Predicting customer churn, which had been conducted as a sub-research area for CRM, has recently become more important as a big data-based performance marketing theme due to the development of business machine learning technology. Until now, research on customer churn prediction has been carried out actively in such sectors as the mobile telecommunication industry, the financial industry, the distribution industry, and the game industry, which are highly competitive and urgent to manage churn. In addition, These churn prediction studies were focused on improving the performance of the churn prediction model itself, such as simply comparing the performance of various models, exploring features that are effective in forecasting departures, or developing new ensemble techniques, and were limited in terms of practical utilization because most studies considered the entire customer group as a group and developed a predictive model. As such, the main purpose of the existing related research was to improve the performance of the predictive model itself, and there was a relatively lack of research to improve the overall customer churn prediction process. In fact, customers in the business have different behavior characteristics due to heterogeneous transaction patterns, and the resulting churn rate is different, so it is unreasonable to assume the entire customer as a single customer group. Therefore, it is desirable to segment customers according to customer classification criteria, such as loyalty, and to operate an appropriate churn prediction model individually, in order to carry out effective customer churn predictions in heterogeneous industries. Of course, in some studies, there are studies in which customers are subdivided using clustering techniques and applied a churn prediction model for individual customer groups. Although this process of predicting churn can produce better predictions than a single predict model for the entire customer population, there is still room for improvement in that clustering is a mechanical, exploratory grouping technique that calculates distances based on inputs and does not reflect the strategic intent of an entity such as loyalties. This study proposes a segment-based customer departure prediction process (CCP/2DL: Customer Churn Prediction based on Two-Dimensional Loyalty segmentation) based on two-dimensional customer loyalty, assuming that successful customer churn management can be better done through improvements in the overall process than through the performance of the model itself. CCP/2DL is a series of churn prediction processes that segment two-way, quantitative and qualitative loyalty-based customer, conduct secondary grouping of customer segments according to churn patterns, and then independently apply heterogeneous churn prediction models for each churn pattern group. Performance comparisons were performed with the most commonly applied the General churn prediction process and the Clustering-based churn prediction process to assess the relative excellence of the proposed churn prediction process. The General churn prediction process used in this study refers to the process of predicting a single group of customers simply intended to be predicted as a machine learning model, using the most commonly used churn predicting method. And the Clustering-based churn prediction process is a method of first using clustering techniques to segment customers and implement a churn prediction model for each individual group. In cooperation with a global NGO, the proposed CCP/2DL performance showed better performance than other methodologies for predicting churn. This churn prediction process is not only effective in predicting churn, but can also be a strategic basis for obtaining a variety of customer observations and carrying out other related performance marketing activities.

• Journal of Intelligence and Information Systems
• /
• v.27 no.4
• /
• pp.1-22
• /
• 2021

Aspect Based Sentiment Analysis (ABSA), which analyzes sentiment based on aspects that appear in the text, is drawing attention because it can be used in various business industries. ABSA is a study that analyzes sentiment by aspects for multiple aspects that a text has. It is being studied in various forms depending on the purpose, such as analyzing all targets or just aspects and sentiments. Here, the aspect refers to the property of a target, and the target refers to the text that causes the sentiment. For example, for restaurant reviews, you could set the aspect into food taste, food price, quality of service, mood of the restaurant, etc. Also, if there is a review that says, "The pasta was delicious, but the salad was not," the words "steak" and "salad," which are directly mentioned in the sentence, become the "target." So far, in ABSA, most studies have analyzed sentiment only based on aspects or targets. However, even with the same aspects or targets, sentiment analysis may be inaccurate. Instances would be when aspects or sentiment are divided or when sentiment exists without a target. For example, sentences like, "Pizza and the salad were good, but the steak was disappointing." Although the aspect of this sentence is limited to "food," conflicting sentiments coexist. In addition, in the case of sentences such as "Shrimp was delicious, but the price was extravagant," although the target here is "shrimp," there are opposite sentiments coexisting that are dependent on the aspect. Finally, in sentences like "The food arrived too late and is cold now." there is no target (NULL), but it transmits a negative sentiment toward the aspect "service." Like this, failure to consider both aspects and targets - when sentiment or aspect is divided or when sentiment exists without a target - creates a dual dependency problem. To address this problem, this research analyzes sentiment by considering both aspects and targets (Target-Aspect-Sentiment Detection, hereby TASD). This study detected the limitations of existing research in the field of TASD: local contexts are not fully captured, and the number of epochs and batch size dramatically lowers the F1-score. The current model excels in spotting overall context and relations between each word. However, it struggles with phrases in the local context and is relatively slow when learning. Therefore, this study tries to improve the model's performance. To achieve the objective of this research, we additionally used auxiliary loss in aspect-sentiment classification by constructing CNN(Convolutional Neural Network) layers parallel to existing models. If existing models have analyzed aspect-sentiment through BERT encoding, Pooler, and Linear layers, this research added CNN layer-adaptive average pooling to existing models, and learning was progressed by adding additional loss values for aspect-sentiment to existing loss. In other words, when learning, the auxiliary loss, computed through CNN layers, allowed the local context to be captured more fitted. After learning, the model is designed to do aspect-sentiment analysis through the existing method. To evaluate the performance of this model, two datasets, SemEval-2015 task 12 and SemEval-2016 task 5, were used and the f1-score increased compared to the existing models. When the batch was 8 and epoch was 5, the difference was largest between the F1-score of existing models and this study with 29 and 45, respectively. Even when batch and epoch were adjusted, the F1-scores were higher than the existing models. It can be said that even when the batch and epoch numbers were small, they can be learned effectively compared to the existing models. Therefore, it can be useful in situations where resources are limited. Through this study, aspect-based sentiments can be more accurately analyzed. Through various uses in business, such as development or establishing marketing strategies, both consumers and sellers will be able to make efficient decisions. In addition, it is believed that the model can be fully learned and utilized by small businesses, those that do not have much data, given that they use a pre-training model and recorded a relatively high F1-score even with limited resources.

• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.155-174
• /
• 2022

From January 2020 to October 2021, more than 500,000 academic studies related to COVID-19 (Coronavirus-2, a fatal respiratory syndrome) have been published. The rapid increase in the number of papers related to COVID-19 is putting time and technical constraints on healthcare professionals and policy makers to quickly find important research. Therefore, in this study, we propose a method of extracting useful information from text data of extensive literature using LDA and Word2vec algorithm. Papers related to keywords to be searched were extracted from papers related to COVID-19, and detailed topics were identified. The data used the CORD-19 data set on Kaggle, a free academic resource prepared by major research groups and the White House to respond to the COVID-19 pandemic, updated weekly. The research methods are divided into two main categories. First, 41,062 articles were collected through data filtering and pre-processing of the abstracts of 47,110 academic papers including full text. For this purpose, the number of publications related to COVID-19 by year was analyzed through exploratory data analysis using a Python program, and the top 10 journals under active research were identified. LDA and Word2vec algorithm were used to derive research topics related to COVID-19, and after analyzing related words, similarity was measured. Second, papers containing 'vaccine' and 'treatment' were extracted from among the topics derived from all papers, and a total of 4,555 papers related to 'vaccine' and 5,971 papers related to 'treatment' were extracted. did For each collected paper, detailed topics were analyzed using LDA and Word2vec algorithms, and a clustering method through PCA dimension reduction was applied to visualize groups of papers with similar themes using the t-SNE algorithm. A noteworthy point from the results of this study is that the topics that were not derived from the topics derived for all papers being researched in relation to COVID-19 (