• The Journal of Society for e-Business Studies
• /
• v.19 no.2
• /
• pp.95-108
• /
• 2014

Opinion Mining is one of the application domains of text mining which extracting opinions from documents, and much researches are currently underway. Most of related researches focused on the sentiment classification which classifies the documents into positive/negative opinions. However, there is a little interest in extracting the features characterizing the individual product. In this paper, we propose the technique classifying the opinion documents according to the product features, and selecting the those features characterizing each product. In the proposed method, we utilize the document clustering technique and develope a new algorithm for evaluating the similarity between documents. In addition, through experiments, we prove the usefulness of proposed method.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.7
• /
• pp.548-553
• /
• 2004

In this paper we propose modified LBG algorithm to minimize quantization errors. When we apply conventional LBG algorithm for speaker verification system, problems that result from small amount of training data can be generated. That is, quantization error comes from fixed-sized codebook without any consideration for speaker characteristics and splitting vector in the wrong direction worsen performance of speaker verification system. So, we propose modified clustering method that has variable sized codebook according to speaker characteristics and makes right splitting direction by finding the farthest member away from mean and then find another member from the member. Simulation results show effectiveness of the proposed algorithm.

• Journal of the Korean Society for Library and Information Science
• /
• v.37 no.2
• /
• pp.89-105
• /
• 2003

In this paper, an effective method for clustering terminologies extracted from text is proposed, in order to develope a search engine to extract relevant information from large web documents. To prevent frequency of the meaningless association rules among general terminologies, only useful association rules among terminologies are produced using database tables which consist of domain-specific terminologies. Such association rules are produced by applying the Apriori algorithm after forming transaction units from groups of association rules in a document. A group of association rules produced from a terminology forms in a cluster.

• Journal of East-West Nursing Research
• /
• v.29 no.2
• /
• pp.185-195
• /
• 2023

Purpose: The purpose of this study is to analyze changes of child welfare policies and provide insights based on the collection and classification of newspaper articles. Methods: Articles related to child welfare policies were collected from 1990, during the Kim, Young-sam administration, to May 9, 2022, under the Moon, Jae-in administration. K-Means clustering and keyword Term Frequency-Inverse Document Frequency analysis were utilized to cluster and analyze newspaper articles with similar themes. Results: The administrations of Kim, Young-sam, Kim, Dae-jung, Roh, Moo-hyun, and Park, Geun-hye were classified into two clusters, and the Lee, Myung-bak and Moon, Jae-in administrations were classified into three clusters. Conclusion: South Korea's child welfare policies have focused on ensuring the safety and healthy development of children through diverse policies initiatives over the years. However, challenges related to child protection and child abuse persist. This requires additional resources and budget allocation. It is important to establish a comprehensive support system for children and families, including comprehensive nursing support.

• The Journal of the Acoustical Society of Korea
• /
• v.20 no.4E
• /
• pp.31-38
• /
• 2001

A large-scale synthesis database for a unit selection based synthesis method usually retains redundant synthesis unit instances, which are useless to the synthetic speech quality. In this paper, to eliminate those instances from the synthesis database, we proposed a new pruning method called weighted vector quantization (WVQ). The WVQ reflects relative importance of each synthesis unit instance when clustering the similar instances using vector quantization (VQ) technique. The proposed method was compared with two conventional pruning methods through the objective and subjective evaluations of the synthetic speech quality: one to simply limit maximum number of instance, and the other based on normal VQ-based clustering. The proposed method showed the best performance under 50% reduction rates. Over 50% of reduction rates, the synthetic speech quality is not seriously but perceptibly degraded. Using the proposed method, the synthesis database can be efficiently reduced without serious degradation of the synthetic speech quality.

• ETRI Journal
• /
• v.44 no.4
• /
• pp.599-612
• /
• 2022

This study focuses on improving a word embedding model to enhance the performance of downstream tasks, such as those of dialog systems. To improve traditional word embedding models, such as skip-gram, it is critical to refine the word features and expand the context model. In this paper, we approach the word model from the perspective of subword embedding and attempt to extend the context model by integrating various sentence models. Our proposed sentence model is a subword-based skip-thought model that integrates self-attention and relative position encoding techniques. We also propose a clustering-based dialog model for downstream task verification and evaluate its relationship with the sentence-model-based subword embedding technique. The proposed subword embedding method produces better results than previous methods in evaluating word and sentence similarity. In addition, the downstream task verification, a clustering-based dialog system, demonstrates an improvement of up to 4.86% over the results of FastText in previous research.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.3
• /
• pp.619-630
• /
• 2015

The purpose of this study is to provide basic information to develop a game strategy plan of a team in a future by identifying the patterns of attack and pass of national men's professional volleyball teams and extracting core key words related with volleyball game performance to evaluate game performance using 'social network analysis' and 'text mining'. As for the analysis result of 'social network analysis' with the whole data, group '0' (6 players) and group '1' (11 players) were partitioned. A point of view the degree centrality and betweenness centrality in 'social network analysis' results, we can know that the group '1' more active game performance than the group '0'. The significant result for two group (win and loss) obtained by 'text mining' according to two groups ('0' and '1') obtained by 'social network analysis' showed significant difference (p-value: 0.001). As for clustering of each network, group '0' had the tendency to score points through set player D and E. In group '1', the player K had the tendency to fail if he attack through 'dig'; players C and D have a good performance through 'set' play.

• Journal of Digital Convergence
• /
• v.17 no.1
• /
• pp.9-27
• /
• 2019

The purpose of this study is to examine consumers' responses to mobile payment services by using a text-mining technique focusing on Samsung Pay as it is used in both online and offline transactions. We conducted text frequency analysis, text clustering analysis, and text network analysis using R programming. The major findings are as follows. First, the most frequently used key words referenced the brand names of the mobile devices, the replacement of traditional wallets and unique functions of Samsung Pay. Second, there was a clear split between positive and negative responses at the macro level. Third, replacement of traditional wallets played a great role in the positive responses and continuous use of mobile payment services. This study provides in-depth understanding of consumer responses toward mobile payment services. It also offers practical implications that may help mobile payment marketers correspond to consumer values and expectations, thus increasing consumer satisfaction.

• Journal of Korean Society for Geospatial Information Science
• /
• v.23 no.2
• /
• pp.89-96
• /
• 2015

Location-based social media data have high potential to be used in various area such as big data, location based services and so on. In this study, we applied a series of analysis methodology to figure out how the important keywords in location-based social media are spatially distributed by analyzing text information. For this purpose, we collected tweet data with geo-tag in Gangnam district and its environs in Seoul for a month of August 2013. From this tweet data, principle keywords are extracted. Among these, keywords of three categories such as food, entertainment and work and study are selected and classified by category. The spatial clustering is conducted to the tweet data which contains keywords in each category. Clusters of each category are compared with buildings and benchmark POIs in the same position. As a result of comparison, clusters of food category showed high consistency with commercial areas of large scale. Clusters of entertainment category corresponded with theaters and sports complex. Clusters of work and study showed high consistency with areas where private institutes and office buildings are concentrated.

• Proceedings of the IEEK Conference
• /
• 2001.06c
• /
• pp.141-144
• /
• 2001

In this paper, a new algorithm for extracting and recognizing characters from video, without pre-knowledge such as font, color, size of character, is proposed. To improve the recognition rate for videos with complex background at low resolution, continuous frames with identical text region are automatically detected to compose an average frame. Using boundary pixels of a text region as seeds, we apply region filling to remove background from the character Then color clustering is applied to remove remaining backgrounds according to the verification of region filling process. Features such as white run and zero-one transition from the center, are extracted from unknown characters. These feature are compared with a pre-composed character feature set to recognize the characters.