• Asia pacific journal of information systems
• /
• v.21 no.2
• /
• pp.89-116
• /
• 2011

In collaborative tagging systems such as Delicious.com and Flickr.com, users assign keywords or tags to their uploaded resources, such as bookmarks and pictures, for their future use or sharing purposes. The collection of resources and tags generated by a user is called a personomy, and the collection of all personomies constitutes the folksonomy. The most significant need of the folksonomy users Is to efficiently find useful resources or experts on specific topics. An excellent ranking algorithm would assign higher ranking to more useful resources or experts. What resources are considered useful In a folksonomic system? Does a standard superior to frequency or freshness exist? The resource recommended by more users with mere expertise should be worthy of attention. This ranking paradigm can be implemented through a graph-based ranking algorithm. Two well-known representatives of such a paradigm are Page Rank by Google and HITS(Hypertext Induced Topic Selection) by Kleinberg. Both Page Rank and HITS assign a higher evaluation score to pages linked to more higher-scored pages. HITS differs from PageRank in that it utilizes two kinds of scores: authority and hub scores. The ranking objects of these pages are limited to Web pages, whereas the ranking objects of a folksonomic system are somewhat heterogeneous(i.e., users, resources, and tags). Therefore, uniform application of the voting notion of PageRank and HITS based on the links to a folksonomy would be unreasonable, In a folksonomic system, each link corresponding to a property can have an opposite direction, depending on whether the property is an active or a passive voice. The current research stems from the Idea that a graph-based ranking algorithm could be applied to the folksonomic system using the concept of mutual Interactions between entitles, rather than the voting notion of PageRank or HITS. The concept of mutual interactions, proposed for ranking the Semantic Web resources, enables the calculation of importance scores of various resources unaffected by link directions. The weights of a property representing the mutual interaction between classes are assigned depending on the relative significance of the property to the resource importance of each class. This class-oriented approach is based on the fact that, in the Semantic Web, there are many heterogeneous classes; thus, applying a different appraisal standard for each class is more reasonable. This is similar to the evaluation method of humans, where different items are assigned specific weights, which are then summed up to determine the weighted average. We can check for missing properties more easily with this approach than with other predicate-oriented approaches. A user of a tagging system usually assigns more than one tags to the same resource, and there can be more than one tags with the same subjectivity and objectivity. In the case that many users assign similar tags to the same resource, grading the users differently depending on the assignment order becomes necessary. This idea comes from the studies in psychology wherein expertise involves the ability to select the most relevant information for achieving a goal. An expert should be someone who not only has a large collection of documents annotated with a particular tag, but also tends to add documents of high quality to his/her collections. Such documents are identified by the number, as well as the expertise, of users who have the same documents in their collections. In other words, there is a relationship of mutual reinforcement between the expertise of a user and the quality of a document. In addition, there is a need to rank entities related more closely to a certain entity. Considering the property of social media that ensures the popularity of a topic is temporary, recent data should have more weight than old data. We propose a comprehensive folksonomy ranking framework in which all these considerations are dealt with and that can be easily customized to each folksonomy site for ranking purposes. To examine the validity of our ranking algorithm and show the mechanism of adjusting property, time, and expertise weights, we first use a dataset designed for analyzing the effect of each ranking factor independently. We then show the ranking results of a real folksonomy site, with the ranking factors combined. Because the ground truth of a given dataset is not known when it comes to ranking, we inject simulated data whose ranking results can be predicted into the real dataset and compare the ranking results of our algorithm with that of a previous HITS-based algorithm. Our semantic ranking algorithm based on the concept of mutual interaction seems to be preferable to the HITS-based algorithm as a flexible folksonomy ranking framework. Some concrete points of difference are as follows. First, with the time concept applied to the property weights, our algorithm shows superior performance in lowering the scores of older data and raising the scores of newer data. Second, applying the time concept to the expertise weights, as well as to the property weights, our algorithm controls the conflicting influence of expertise weights and enhances overall consistency of time-valued ranking. The expertise weights of the previous study can act as an obstacle to the time-valued ranking because the number of followers increases as time goes on. Third, many new properties and classes can be included in our framework. The previous HITS-based algorithm, based on the voting notion, loses ground in the situation where the domain consists of more than two classes, or where other important properties, such as "sent through twitter" or "registered as a friend," are added to the domain. Forth, there is a big difference in the calculation time and memory use between the two kinds of algorithms. While the matrix multiplication of two matrices, has to be executed twice for the previous HITS-based algorithm, this is unnecessary with our algorithm. In our ranking framework, various folksonomy ranking policies can be expressed with the ranking factors combined and our approach can work, even if the folksonomy site is not implemented with Semantic Web languages. Above all, the time weight proposed in this paper will be applicable to various domains, including social media, where time value is considered important.

• Asia pacific journal of information systems
• /
• v.20 no.1
• /
• pp.81-99
• /
• 2010

Lately, in consumers' markets the number of new items is rapidly increasing at an overwhelming rate while consumers have limited access to information about those new products in making a sensible, well-informed purchase. Therefore, item providers and customers need a system which recommends right items to right customers. Also, whenever new items are released, for instance, the recommender system specializing in new items can help item providers locate and identify potential customers. Currently, new items are being added to an existing system without being specially noted to consumers, making it difficult for consumers to identify and evaluate new products introduced in the markets. Most of previous approaches for recommender systems have to rely on the usage history of customers. For new items, this content-based (CB) approach is simply not available for the system to recommend those new items to potential consumers. Although collaborative filtering (CF) approach is not directly applicable to solve the new item problem, it would be a good idea to use the basic principle of CF which identifies similar customers, i,e. neighbors, and recommend items to those customers who have liked the similar items in the past. This research aims to suggest a hybrid recommendation procedure based on the preference boundary of target customer. We suggest the hybrid recommendation procedure using the preference boundary in the feature space for recommending new items only. The basic principle is that if a new item belongs within the preference boundary of a target customer, then it is evaluated to be preferred by the customer. Customers' preferences and characteristics of items including new items are represented in a feature space, and the scope or boundary of the target customer's preference is extended to those of neighbors'. The new item recommendation procedure consists of three steps. The first step is analyzing the profile of items, which are represented as k-dimensional feature values. The second step is to determine the representative point of the target customer's preference boundary, the centroid, based on a personal information set. To determine the centroid of preference boundary of a target customer, three algorithms are developed in this research: one is using the centroid of a target customer only (TC), the other is using centroid of a (dummy) big target customer that is composed of a target customer and his/her neighbors (BC), and another is using centroids of a target customer and his/her neighbors (NC). The third step is to determine the range of the preference boundary, the radius. The suggested algorithm Is using the average distance (AD) between the centroid and all purchased items. We test whether the CF-based approach to determine the centroid of the preference boundary improves the recommendation quality or not. For this purpose, we develop two hybrid algorithms, BC and NC, which use neighbors when deciding centroid of the preference boundary. To test the validity of hybrid algorithms, BC and NC, we developed CB-algorithm, TC, which uses target customers only. We measured effectiveness scores of suggested algorithms and compared them through a series of experiments with a set of real mobile image transaction data. We spilt the period between 1st June 2004 and 31st July and the period between 1st August and 31st August 2004 as a training set and a test set, respectively. The training set Is used to make the preference boundary, and the test set is used to evaluate the performance of the suggested hybrid recommendation procedure. The main aim of this research Is to compare the hybrid recommendation algorithm with the CB algorithm. To evaluate the performance of each algorithm, we compare the purchased new item list in test period with the recommended item list which is recommended by suggested algorithms. So we employ the evaluation metric to hit the ratio for evaluating our algorithms. The hit ratio is defined as the ratio of the hit set size to the recommended set size. The hit set size means the number of success of recommendations in our experiment, and the test set size means the number of purchased items during the test period. Experimental test result shows the hit ratio of BC and NC is bigger than that of TC. This means using neighbors Is more effective to recommend new items. That is hybrid algorithm using CF is more effective when recommending to consumers new items than the algorithm using only CB. The reason of the smaller hit ratio of BC than that of NC is that BC is defined as a dummy or virtual customer who purchased all items of target customers' and neighbors'. That is centroid of BC often shifts from that of TC, so it tends to reflect skewed characters of target customer. So the recommendation algorithm using NC shows the best hit ratio, because NC has sufficient information about target customers and their neighbors without damaging the information about the target customers.

• Journal of Distribution Science
• /
• v.16 no.5
• /
• pp.83-90
• /
• 2018

Purpose - Journals of Korea Distribution Science Association (KODISA) made great efforts in responding to the constant shifts in academic paradigms and in producing synergetic effects among KODISA journals to achieve the goal of maintaining their status in the world's reputable scholarly journals. The aim of this study is to analyze the current practice and performance of KODISA journals and develop strategies that will continuously meet and respond to the changes and success in the future. Research design, data, and methodology - This is a case study, an analytical approach, which focuses on analyzing current and previous strategies, practices, and performances of the four major journals of KODISA and the association. The organizational structure, including election and terms of KODISA officers, new membership, and members of editorial board, is discussed and analyzed. The citation, submission, publication, and rejection rates of all four journals are examined, and the progress, including the status of indexing of each journal, is discussed. Results - The analysis indicates that KODISA has significantly invested its resources into improving its journals and attracting new members. The analysis also shows the strategy of the organizational structure, which includes election and terms of officers and editorial board members that implemented over the years, was successful. Both Journal of Distribution Science (JDS) and Journal of Finance, Economics, and Business (JAFEB) are indexed in SCOPUS, with East Asian Journal of Business Management (EAJBM) in the final stage of the SCOPUS indexing evaluation, and International Journal of Industrial Distribution and Business (IJIDB) will complete and submit their indexing evaluation materials to SCOPUS this summer. Conclusions - The success and progress of KODISA and its journals clearly support the need for continuous development, analysis, revision, and implementation of strategies. Based on the analysis, conducting the annual performance reviews of the association and its journals and planning and strategizing based on the reviews since 2011 have greatly contributed to the overall success. In terms of meeting the short term strategy, KODISA has to continue developing relationships with relevant and appropriate scholarly/academic associations to expand the scope of its business, establishing independence of each journal and its respective procedures and practices and improving the quality of the journals and their publications through KODISA's international conferences.

• Strategy21
• /
• s.33
• /
• pp.65-112
• /
• 2014

Recent trends show that the PRC has stepped aside its "army-centered approach" and placed greater emphasis on its Navy and Air Force for a wider range of operations, thereby reducing its ground force and harnessing its economic power and military technology into naval development. A quantitative growth of the PLA Navy itself is no surprise as this is not a recent phenomenon. Now is the time to pay closer attention to the level of PRC naval force's performance and the extent of its warfighting capacity in the maritime domain. It is also worth asking what China can do with its widening naval power foundation. In short, it is time to delve into several possible scenarios I which the PRC poses a real threat. With this in mind, in Section Two the paper seeks to observe the construction progress of PRC's naval power and its future prospects up to the year 2020, and categorize time frame according to its major force improvement trends. By analyzing qualitative improvements made over time, such as the scale of investment and the number of ships compared to increase in displacement (tonnage), this paper attempts to identify salient features in the construction of naval power. Chapter Three sets out performance evaluation on each type of PRC naval ships as well as capabilities of the Navy, Air Force, the Second Artillery (i.e., strategic missile forces) and satellites that could support maritime warfare. Finall, the concluding chapter estimates the PRC's maritime warfighting capability as anticipated in respective conflict scenarios, and considers its impact on the Korean Peninsula and proposes the directions ROK should steer in response. First of all, since the 1980s the PRC navy has undergone transitions as the focus of its military strategic outlook shifted from ground warfare to maritime warfare, and within 30 years of its effort to construct naval power while greatly reducing the size of its ground forces, the PRC has succeeded in building its naval power next to the U.S.'s in the world in terms of number, with acquisition of an aircraft carrier, Chinese-version of the Aegis, submarines and so on. The PRC also enjoys great potentials to qualitatively develop its forces such as indigenous aircraft carriers, next-generation strategic submarines, next-generation destroyers and so forth, which is possible because the PRC has accumulated its independent production capabilities in the process of its 30-year-long efforts. Secondly, one could argue that ROK still has its chances of coping with the PRC in naval power since, despite its continuous efforts, many estimate that the PRC naval force is roughly ten or more years behind that of superpowers such as the U.S., on areas including radar detection capability, EW capability, C4I and data-link systems, doctrines on force employment as well as tactics, and such gap cannot be easily overcome. The most probable scenarios involving the PRC in sea areas surrounding the Korean Peninsula are: first, upon the outbreak of war in the peninsula, the PRC may pursue military intervention through sea, thereby undermining efforts of the ROK-U.S. combined operations; second, ROK-PRC or PRC-Japan conflicts over maritime jurisdiction or ownership over the Senkaku/Diaoyu islands could inflict damage to ROK territorial sovereignty or economic gains. The PRC would likely attempt to resolve the conflict employing blitzkrieg tactics before U.S. forces arrive on the scene, while at the same time delaying and denying access of the incoming U.S. forces. If this proves unattainable, the PRC could take a course of action adopting "long-term attrition warfare," thus weakening its enemy's sustainability. All in all, thiss paper makes three proposals on how the ROK should respond. First, modern warfare as well as the emergent future warfare demonstrates that the center stage of battle is no longer the domestic territory, but rather further away into the sea and space. In this respect, the ROKN should take advantage of the distinct feature of battle space on the peninsula, which is surrounded by the seas, and obtain capabilities to intercept more than 50 percent of the enemy's ballistic missiles, including those of North Korea. In tandem with this capacity, employment of a large scale of UAV/F Carrier for Kill Chain operations should enhance effectiveness. This is because conditions are more favorable to defend from sea, on matters concerning accuracy rates against enemy targets, minimized threat of friendly damage, and cost effectiveness. Second, to maintain readiness for a North Korean crisis where timely deployment of US forces is not possible, the ROKN ought to obtain capabilities to hold the enemy attack at bay while deterring PRC naval intervention. It is also argued that ROKN should strengthen its power so as to protect national interests in the seas surrounding the peninsula without support from the USN, should ROK-PRC or ROK-Japan conflict arise concerning maritime jurisprudence. Third, the ROK should fortify infrastructures for independent construction of naval power and expand its R&D efforts, and for this purpose, the ROK should make the most of the advantages stemming from the ROK-U.S. alliance inducing active support from the United States. The rationale behind this argument is that while it is strategically effective to rely on alliance or jump on the bandwagon, the ultimate goal is always to acquire an independent response capability as much as possible.

• Journal of Intelligence and Information Systems
• /
• v.20 no.2
• /
• pp.123-136
• /
• 2014

Recently, online shopping has further developed as the use of the Internet and a variety of smart mobile devices becomes more prevalent. The increase in the scale of such shopping has led to the creation of many Internet shopping malls. Consequently, there is a tendency for increasingly fierce competition among online retailers, and as a result, many Internet shopping malls are making significant attempts to attract online users to their sites. One such attempt is keyword marketing, whereby a retail site pays a fee to expose its link to potential customers when they insert a specific keyword on an Internet portal site. The price related to each keyword is generally estimated by the keyword's frequency of appearance. However, it is widely accepted that the price of keywords cannot be based solely on their frequency because many keywords may appear frequently but have little relationship to shopping. This implies that it is unreasonable for an online shopping mall to spend a great deal on some keywords simply because people frequently use them. Therefore, from the perspective of shopping malls, a specialized process is required to extract meaningful keywords. Further, the demand for automating this extraction process is increasing because of the drive to improve online sales performance. In this study, we propose a methodology that can automatically extract only shopping-related keywords from the entire set of search keywords used on portal sites. We define a shopping-related keyword as a keyword that is used directly before shopping behaviors. In other words, only search keywords that direct the search results page to shopping-related pages are extracted from among the entire set of search keywords. A comparison is then made between the extracted keywords' rankings and the rankings of the entire set of search keywords. Two types of data are used in our study's experiment: web browsing history from July 1, 2012 to June 30, 2013, and site information. The experimental dataset was from a web site ranking site, and the biggest portal site in Korea. The original sample dataset contains 150 million transaction logs. First, portal sites are selected, and search keywords in those sites are extracted. Search keywords can be easily extracted by simple parsing. The extracted keywords are ranked according to their frequency. The experiment uses approximately 3.9 million search results from Korea's largest search portal site. As a result, a total of 344,822 search keywords were extracted. Next, by using web browsing history and site information, the shopping-related keywords were taken from the entire set of search keywords. As a result, we obtained 4,709 shopping-related keywords. For performance evaluation, we compared the hit ratios of all the search keywords with the shopping-related keywords. To achieve this, we extracted 80,298 search keywords from several Internet shopping malls and then chose the top 1,000 keywords as a set of true shopping keywords. We measured precision, recall, and F-scores of the entire amount of keywords and the shopping-related keywords. The F-Score was formulated by calculating the harmonic mean of precision and recall. The precision, recall, and F-score of shopping-related keywords derived by the proposed methodology were revealed to be higher than those of the entire number of keywords. This study proposes a scheme that is able to obtain shopping-related keywords in a relatively simple manner. We could easily extract shopping-related keywords simply by examining transactions whose next visit is a shopping mall. The resultant shopping-related keyword set is expected to be a useful asset for many shopping malls that participate in keyword marketing. Moreover, the proposed methodology can be easily applied to the construction of special area-related keywords as well as shopping-related ones.

• Journal of Agricultural Extension & Community Development
• /
• v.21 no.2
• /
• pp.1-28
• /
• 2014

Private consulting firms that are currently participating in rural area development projects with a bottom-up approach are involved in nearly all areas of rural area development, and the policy environment that emphasizes the bottom-up approach will further expand their participation. Reviews of private consulting firms, which started out with high expectations in the beginning, are now becoming rather negative. Expertise is the key issue in the controversy over private consulting firms, and the analysis tends to limit the causes of the problems within firms. This study was conducted on the premise that the fixation on cause and structure results in policy issues in the promotion process. That is because the government authorities are responsible for managing and supervising the implementation of policies, not developing the policies. The current issues with consulting firms emerged because of the hasty implementation of private consulting through the government policy trend without sufficient consideration, as well as the policy environment that demanded short-term outcomes even though the purpose of bottom-up rural area development lies in the ideology of endogenous development focused on the changes in residents' perceptions. Research was conducted to determine how the problems of private consulting firms that emerged and were addressed in this context influenced the consulting market, using current data and based on the firms' business performance. In analyzing the types, firms were divided into three groups: top performers including market leaders (9), excellent performers (36), and average performers (34). An analysis of the correlation between the business performance of each type and managerial resources such as each firm's expertise revealed that there was only a correlation between human resources and regional development in excellent performers, and none was found with the other types. These results imply that external factors other than a firm's capabilities (e.g., expertise) play a significant role in the standards of selecting private consulting firms. Thus, government authorities must reflect on their error of hastily adopting private consulting firms without sufficient consideration and must urgently establish response measures.

• Korean Security Journal
• /
• no.42
• /
• pp.7-36
• /
• 2015

This study aims at analyzing difference of social Security network, Community unity and local government trust according to socio-demographical features, exploring the relation of social Security network, Community unity and local government trust according to socio-demographical features, presenting results between each variable as a model and verifying the property of mutual ones. This study sampled general citizens in Gwangju for about 15 days Aug. 15 through Aug. 30, 2014, distributed total 450 copies using cluster random sampling, gathered 438 persons, 412 persons of whom were used for analysis. This study verified the validity and credibility of the questionnaire through an experts' meeting, preliminary test, factor analysis and credibility analysis. The credibility of questionnaire was ${\alpha}=.809{\sim}{\alpha}=.890$. The inout data were analyzed by study purpose using SPSSWIN 18.0, as statistical techniques, factor analysis, credibility analysis, correlation analysis, independent sample t verification, ANOVA, multi-regression analysis, path analysis etc. were used. the findings obtained through the above study methods are as follows. First, building a social Security network has an effect on Community institution. That is, the more activated a, the higher awareness on institution. the more activated street CCTV facilities, anti-crime design, local government Security education, the higher the stability. Second, building a social Security network has an effect on trust of local government. That is, the activated local autonomous anti-crime activity, anti-crime design. local government's Security education, police public oder service, the more increased trust of policy, service management, busines performance. Third, Community unity has an effect on trust of local government. That is, the better Community institution is achieved, the higher trust of policy. Also the stabler Community institution, the higher trust of business performance. Fourth, building a social Security network has a direct or indirect effect on Community unity and local government trust. That is, social Security network has a direct effect on trust of local government, but it has a higher effect through Community unity of parameter. Such results showed that Community unity in Gwangju Region is an important factor, which means it is an important variable mediating building a social Security network and trust of local government. To win trust of local residents, we need to prepare for various cultural events and active communication space and build a social Security network for uniting them.

• Journal of The Korean Association For Science Education
• /
• v.36 no.4
• /
• pp.563-579
• /
• 2016

This study seeks educational implication by analyzing research papers dealing with science assessment in the most recent 30 years in Korea. The main purpose of the study is to analyze the trends in published papers on science assessment, their purpose, methodology, and key words, especially concentrating on the cognitive and affective domains. We selected 273 research articles and categorized them by research object, subject, methodology, and contents. To examine the factors that affect the research trend, we also tried to contextualize papers' theme in terms of changes in national curriculum and assessment system during the contemporary period. As a result, an overall research trend reflects changes in science curriculum and assessment events such as implementation of college scholastic ability test or performance assessment. There is an unequal distribution in various aspects of the researches, showing a superiority in cognitive domains than the affective ones. By using standardized data obtained through the national and international assessment of educational achievement in science, quantitative researches were superior to qualitative ones. Studies on cognitive domain use variously written- and performance-based tests, whereas most studies of the affective ones prefer written tests. Applied research and evaluation research are predominant comparing to basic ones, which most of the research methodology is based on statistics. Lastly, we found out that key words and subjects tend to be subdivided and detailed rather than general and comprehensive, as time goes on. Such trend will be helpful to elaborate and refine assessment tools that have been regarded as a problem.

• Korea Journal of Hospital Management
• /
• v.7 no.1
• /
• pp.64-87
• /
• 2002

The purpose of this study is aimed to grasp the factors, which may influence the harmonized organizational efficiency of the objects of hospital structure as well as its structural constituents of each departments of administration, nursing, and para-medical personnels, in order to provide basic data enable to contribute in the development of hospital. The survey data involved in the study was derived from 385 personnels working in 5 medium and small sized hospitals in Seoul area. The main finding of the study can be summarized as follows; 1. The organizational efficiency in accordance with the general characteristics of subjects in order of male, over 30 years of age, university graduates, long-term tenure and high position is higher, whereas, in as much as wage and well trained personnels in administration departments are higher, its organizational efficiency is higher in comparison with those of nursing and para-medical departments. 2. The organizational efficiency in accordance with satisfaction and the motive contributional factors is higher as much as the high satisfactory level in every departments in general. 3. The organizational efficiency in accordance with the factors of job characteristics is higher in as much as difficulty of the jobs is lesser, however there was not statically significance between administration and para-medical departments. In as much as the job circulation is intact, job standard level is higher and the more job responsibility the higher organizational efficiency, while the more workload and the more work feud resulted lower organizational efficiency. It was obvious that the higher professional expertise as well as the training and application level are improving the organizational efficiency. 4. The organizational efficiency in accordance with the factors of structural characteristics was higher in as much as the intercommunication was smooth and the structural formalization level are higher, however there was not statically significance between the participation level of decision making and the organizational efficiency. 5. In as much as higher educational level of over university graduates, management of organization and the job level are satisfied, the psychological motive contributional level is higher, while the lesser job difficulty, the smooth job performance, the higher level of professional expertise, the higher structural formalization level, the smooth intercommunication, have affected as major influence factors of the structural characteristics of organizational efficiency. 6. As the management of hospital organization, the job level and personal relation are satisfied or psychological motive is provided, especially when there are no difficult jobs or smooth job circulation and no job feud are prevailing, it was apparent that the organizational efficiency is improving accordingly. The nursing departments has high educational standard and is satisfied in the management and job level of hospital organization as there are no difficult jobs while the level of hospital's organizational formality is high and the intercommunication is smooth, which are improving the organizational efficiency. The para-medical departments is also satisfied the management and job level of hospital organization and it was apparent that the organizational efficiency is higher in as much as the level of job standardization is high and the intercommunication is smooth. As a result of this study, in order for improving the organizational efficiency of the medium and small sized hospitals, the management and job level as well as personal relation are preferably satisfied, whereas the level of job circulation, job responsibility, the expertise and formalization of organization, intercommunication and etc. should be satisfied, and, therefore, it is advisable to buildup discriminated organizational management and environment for different division on the basis above factors. Since this study is carried on several hospitals in Seoul area, there is a certain limit to generalize its result to all domestic hospitals, nevertheless the gallop poll was made by developing the questionnaires with reasonability and reliability. Especially, as the study was carried by analyzing the comparison of influence factors' difference of organizational efficiency in accordance with the divisional characteristics of the medium and small sized hospitals.

• Journal of Intelligence and Information Systems
• /
• v.20 no.3
• /
• pp.77-92
• /
• 2014

Recently, numerous documents including unstructured data and text have been created due to the rapid increase in the usage of social media and the Internet. Each document is usually provided with a specific category for the convenience of the users. In the past, the categorization was performed manually. However, in the case of manual categorization, not only can the accuracy of the categorization be not guaranteed but the categorization also requires a large amount of time and huge costs. Many studies have been conducted towards the automatic creation of categories to solve the limitations of manual categorization. Unfortunately, most of these methods cannot be applied to categorizing complex documents with multiple topics because the methods work by assuming that one document can be categorized into one category only. In order to overcome this limitation, some studies have attempted to categorize each document into multiple categories. However, they are also limited in that their learning process involves training using a multi-categorized document set. These methods therefore cannot be applied to multi-categorization of most documents unless multi-categorized training sets are provided. To overcome the limitation of the requirement of a multi-categorized training set by traditional multi-categorization algorithms, we propose a new methodology that can extend a category of a single-categorized document to multiple categorizes by analyzing relationships among categories, topics, and documents. First, we attempt to find the relationship between documents and topics by using the result of topic analysis for single-categorized documents. Second, we construct a correspondence table between topics and categories by investigating the relationship between them. Finally, we calculate the matching scores for each document to multiple categories. The results imply that a document can be classified into a certain category if and only if the matching score is higher than the predefined threshold. For example, we can classify a certain document into three categories that have larger matching scores than the predefined threshold. The main contribution of our study is that our methodology can improve the applicability of traditional multi-category classifiers by generating multi-categorized documents from single-categorized documents. Additionally, we propose a module for verifying the accuracy of the proposed methodology. For performance evaluation, we performed intensive experiments with news articles. News articles are clearly categorized based on the theme, whereas the use of vulgar language and slang is smaller than other usual text document. We collected news articles from July 2012 to June 2013. The articles exhibit large variations in terms of the number of types of categories. This is because readers have different levels of interest in each category. Additionally, the result is also attributed to the differences in the frequency of the events in each category. In order to minimize the distortion of the result from the number of articles in different categories, we extracted 3,000 articles equally from each of the eight categories. Therefore, the total number of articles used in our experiments was 24,000. The eight categories were "IT Science," "Economy," "Society," "Life and Culture," "World," "Sports," "Entertainment," and "Politics." By using the news articles that we collected, we calculated the document/category correspondence scores by utilizing topic/category and document/topics correspondence scores. The document/category correspondence score can be said to indicate the degree of correspondence of each document to a certain category. As a result, we could present two additional categories for each of the 23,089 documents. Precision, recall, and F-score were revealed to be 0.605, 0.629, and 0.617 respectively when only the top 1 predicted category was evaluated, whereas they were revealed to be 0.838, 0.290, and 0.431 when the top 1 - 3 predicted categories were considered. It was very interesting to find a large variation between the scores of the eight categories on precision, recall, and F-score.