• 말소리와 음성과학
• /
• 제13권3호
• /
• pp.65-70
• /
• 2021

This paper proposes a method for designing a large recording script for open domain English speech synthesis. For read-aloud style text, 12 domains and 294 sub-domains were designed using text contained in five different news media publications. For conversational style text, 4 domains and 36 sub-domains were designed using movie subtitles. The final script consists of 43,013 sentences, 27,085 read-aloud style sentences, and 15,928 conversational style sentences, consisting of 549,683 tokens and 38,356 types. The completed script is analyzed using four criteria: word coverage (type coverage and token coverage), high-frequency vocabulary coverage, phonetic coverage (diphone coverage and triphone coverage), and readability. The type coverage of our script reaches 36.86% despite its low token coverage of 2.97%. The high-frequency vocabulary coverage of the script is 73.82%, and the diphone coverage and triphone coverage of the whole script is 86.70% and 38.92%, respectively. The average readability of whole sentences is 9.03. The results of analysis show that the proposed method is effective in producing a large recording script for English speech synthesis, demonstrating good coverage in terms of unique words, high-frequency vocabulary, phonetic units, and readability.

• 한국의류학회지
• /
• 제47권1호
• /
• pp.35-51
• /
• 2023

This study used text mining to extract 2,842 keywords from 7,397 product names and organized them into categories in order to analyze the characteristics of keywords appearing in the product names of jeans after 2020. The item category included denim and Chungbaji [청바지], and Ilja [일자], while the silhouette category included wide and bootcut. In addition, high-waist and banding comprised the making sector, and the materials category consisted of napping, spandex, and soft blue. Denim surpassed the others in frequency, co-occurrence frequency, and centrality, and co-appeared with various other keywords. Also, the co-appearance of item and silhouette was prominent, and there were many keyword combinations that showed characteristics related to (a) high waist; (b) hemline detail; (c) rubber band; and (d) partial tearing. Furthermore, idiom expressions such as 'slim fit' and 'back tearing', which were not highlighted in the co-occurrence frequency, were additionally confirmed through correlation. Therefore, the product name analysis effectively identified the detailed characteristics of the silhouette and the making of jeans preferred by consumers.

• 정보처리학회논문지B
• /
• 제10B권7호
• /
• pp.725-734
• /
• 2003

본 논문에서는 통계적 분석 기법인 주성분 분석과 비정칙치 분해를 이용한 문서 방법을 제안한다. 제안한 방법은 문서내의 주제어를 추출한 후, 추출된 주제어와 문장간의 거리가 가장 짧은 문장들을 중요 문장으로 추출하여 요약으로 제시한다. 주제어를 추출하기 위해서는 주성분 분석을 이용하였으며, 이는 문서 자체내의 빈도 정보와 단어간의 연관 정보를 이용한 것이다. 그리고, 중요 문장을 추출하기 위해 비정칙치 분해를 시행하여 문장 벡터와 주제어 벡터론 획득한 후, 두 벡터간의 유클리디언 거리를 계산하였다. 신문 기사를 대상으로 실험한 결과, 제안한 방법이 출현 빈도만을 이용한 방법과 주성분 분석만을 이용한 방법보다 성능이 우수함을 알 수 있었다.

• Journal of Information Processing Systems
• /
• 제17권4호
• /
• pp.818-833
• /
• 2021

To address the problems of low precision rate, insufficient feature extraction, and poor contextual ability in existing text sentiment analysis methods, a mixed model account of a CNN-BiLSTM-TE (convolutional neural network, bidirectional long short-term memory, and topic extraction) model was proposed. First, Chinese text data was converted into vectors through the method of transfer learning by Word2Vec. Second, local features were extracted by the CNN model. Then, contextual information was extracted by the BiLSTM neural network and the emotional tendency was obtained using softmax. Finally, topics were extracted by the term frequency-inverse document frequency and K-means. Compared with the CNN, BiLSTM, and gate recurrent unit (GRU) models, the CNN-BiLSTM-TE model's F1-score was higher than other models by 0.0147, 0.006, and 0.0052, respectively. Then compared with CNN-LSTM, LSTM-CNN, and BiLSTM-CNN models, the F1-score was higher by 0.0071, 0.0038, and 0.0049, respectively. Experimental results showed that the CNN-BiLSTM-TE model can effectively improve various indicators in application. Lastly, performed scalability verification through a takeaway dataset, which has great value in practical applications.

• 응용통계연구
• /
• 제35권1호
• /
• pp.147-159
• /
• 2022

오늘날 다양한 소셜 미디어 플랫폼이 널리 퍼져 있고 사람들은 그들의 일상생활 속에서 밀접하게 그러한 플랫폼들을 이용하고 있다. 이에 따라, 많은 수의 구독자, 시청, 댓글 등을 보유한 인플루언서들은 우리 사회 속에서 큰 영향력을 가지게 되었다. 이러한 추세에 따라 많은 회사들은 그들의 상품과 서비스 판매의 촉진을 위한 마케팅 목적으로 인플루언서들을 적극 활용하고 있다. 본 연구에서는 게임 마케팅을 위한 비디오에서 인플루언서들의 스피치를 추출하고 텍스트화하여 이를 텍스트 마이닝 기술을 이용하여 탐색적으로 분석한다. 분석에 있어, 성공한 마케팅 비디오와 실패한 마케팅 비디오를 구분하고 성공, 실패한 마케팅 비디오에서 인플루언서들의 언어적 특징들을 비교 분석한다.

• ETRI Journal
• /
• 제43권1호
• /
• pp.95-108
• /
• 2021

Application (app) ratings are feedback provided voluntarily by users and serve as important evaluation criteria for apps. However, these ratings can often be biased owing to insufficient or missing votes. Additionally, significant differences have been observed between numeric ratings and user reviews. This study aims to predict the numeric ratings of Google apps using machine learning classifiers. It exploits numeric app ratings provided by users as training data and returns authentic mobile app ratings by analyzing user reviews. An ensemble learning model is proposed for this purpose that considers term frequency/inverse document frequency (TF/IDF) features. Three TF/IDF features, including unigrams, bigrams, and trigrams, were used. The dataset was scraped from the Google Play store, extracting data from 14 different app categories. Biased and unbiased user ratings were discriminated using TextBlob analysis to formulate the ground truth, from which the classifier prediction accuracy was then evaluated. The results demonstrate the high potential for machine learning-based classifiers to predict authentic numeric ratings based on actual user reviews.

• 한국의류학회지
• /
• 제46권1호
• /
• pp.116-131
• /
• 2022

This study used text mining to analyze big data to understand consumers' demand for and perceptions of fashion masks. Based on the text-mining analysis results, a survey was conducted with those living in Korea to investigate the influence of consumers' mask selection criteria on mask brand awareness and purchase intention for fashion masks. "Fashion mask" and "functional mask" were used as the keywords in a text-mining analysis, and an online survey of 242 respondents was conducted. The analysis results were as follows: First, the text-mining analysis extracted commonly appearing words that had a high frequency and TF-IDF, such as "COVID-19," "fashion," "celebrity," "antibacterial," and "filter." This confirmed that during the COVID-19 pandemic, consumers have demanded masks that are both functional and fashionable. Second, among consumers' mask selection criteria, trend and design had positive effects on face-mask brand awareness. Third, face-mask brand awareness had a positive effect on the purchase intention for both brand and fashion masks, and the purchase intention for brand masks had a positive effect on the purchase intention for fashion masks.

• 공학교육연구
• /
• 제25권5호
• /
• pp.85-93
• /
• 2022

Capstone Design, a project-based learning technique, is the most important curriculum that clarifying major knowledge and cultivating the ability to apply through the process of solving problems in the industrial field centered on the student project team. Accordingly, various and extensive studies are being conducted for the successful implementation of capstone design courses. Unlike previous studies, this study aimed to quantitatively analyze the opinions that recorded the experiences and feelings of students who performed capstone design, and used text mining methodologies such as frequency analysis, correlation analysis, topic modeling, and sentiment analysis. As a result of examining the overall opinions of the latter period through frequency analysis and correlation analysis, there was a difference between the languages used by the students in the opinions according to gender and project results. Through topic modeling analysis, 'topic selection' and 'the relationship between team members' showed an increase in occupancy or high occupancy, and topics such as 'presentation', 'leadership', and 'feeling what they felt' showed a tendency to decreasing occupancy. Lastly, sentiment analysis has found that female students showed more neutral emotions than male students, and the passed group showed more negative emotions than the non-passed group and less neutral emotions. Based on these findings, students' practical recognition of the curriculum was considered and implications for the improvement of capstone design were presented.

• 한국IT서비스학회지
• /
• 제15권3호
• /
• pp.71-95
• /
• 2016

In recent years, text mining has been used to extract meaningful insights from the large volume of unstructured text data sets of various domains. As one of the most representative text mining applications, topic modeling has been widely used to extract main topics in the form of a set of keywords extracted from a large collection of documents. In general, topic modeling is performed according to the weighted frequency of words in a document corpus. However, general topic modeling cannot discover the relation between documents if the documents share only a few terms, although the documents are in fact strongly related from a particular perspective. For instance, a document about "sexual offense" and another document about "silver industry for aged persons" might not be classified into the same topic because they may not share many key terms. However, these two documents can be strongly related from the R&D perspective because some technologies, such as "RF Tag," "CCTV," and "Heart Rate Sensor," are core components of both "sexual offense" and "silver industry." Thus, in this study, we attempted to discover the differences between the results of general topic modeling and R&D perspective topic modeling. Furthermore, we package social issues from the R&D perspective and present a prototype system, which provides a package of news articles for each R&D issue. Finally, we analyze the quality of R&D perspective topic modeling and provide the results of inter- and intra-topic analysis.

• International Journal of Computer Science & Network Security
• /
• 제22권8호
• /
• pp.87-96
• /
• 2022

This paper employs a computer-aided text analysis (CATA) and a Critical Discourse Analysis (CDA) to explore the strategies of recruitment and intellectual polarization in ISIS (Islamic State in Iraq and Syria) media. The paper's main objective is to shed light on the efficacy of employing computer software in the linguistic analysis of texts, and the extent to which CATA software contribute to deciphering hidden meanings of texts as well as to arrive at concise and authentic results from these texts. More specifically, this paper attempts to demonstrate the contribution of CATA software represented in the two variables of Frequency Distribution Analysis (FDA) and Content Analysis (CA) in decoding the strategies of recruitment and intellectual polarization in one of ISIS 's digital publication: Rumiyah (a digital magazine published by ISIS). The analytical focus is on three strategies of recruitment and intellectual polarization: (i) lexicalization, (ii) intertextual religionisation, and (iii) justification. Two main findings are revealed in this study. First, the application of CATA software into the linguistic investigation of texts contributes effectively to the understanding of the thematic and ideological messages pertaining to the analyzed text. Second, the computational analysis guarantees concise, credible, authentic and ample results than is the case if the analysis is conducted without the work of computer software. The paper, therefore, recommends the integration of CATA software into the linguistic analysis of the various types of texts.