• Journal of Information Science Theory and Practice
• /
• v.10 no.3
• /
• pp.40-56
• /
• 2022

This paper aims to implement a topic modeling technique for extracting the topics of online discussions among library professionals in India. Topic modeling is the established text mining technique popularly used for modeling text data from Twitter, Facebook, Yelp, and other social media platforms. The present study modeled the online discussions of Library and Information Science (LIS) professionals posted on Lis Links. The text data of these posts was extracted using a program written in R using the package "rvest." The data was pre-processed to remove blank posts, posts having text in non-English fonts, punctuation, URLs, emails, etc. Topic modeling with the Latent Dirichlet Allocation algorithm was applied to the pre-processed corpus to identify each topic associated with the posts. The frequency analysis of the occurrence of words in the text corpus was calculated. The results found that the most frequent words included: library, information, university, librarian, book, professional, science, research, paper, question, answer, and management. This shows that the LIS professionals actively discussed exams, research, and library operations on the forum of Lis Links. The study categorized the online discussions on Lis Links into ten topics, i.e. "LIS Recruitment," "LIS Issues," "Other Discussion," "LIS Education," "LIS Research," "LIS Exams," "General Information related to Library," "LIS Admission," "Library and Professional Activities," and "Information Communication Technology (ICT)." It was found that the majority of the posts belonged to "LIS Exam," followed by "Other Discussions" and "General Information related to the Library."

• Journal of KIISE
• /
• v.41 no.9
• /
• pp.652-665
• /
• 2014

Many important terminologies in biomedical text are expressed as abbreviations or acronyms. We newly suggest a semantic link topic model based on the concepts of topic and dependency link to disambiguate biomedical abbreviations and cluster long form variants of abbreviations which refer to the same senses. This model is a generative model inspired by the latent Dirichlet allocation (LDA) topic model, in which each document is viewed as a mixture of topics, with each topic characterized by a distribution over words. Thus, words of a document are generated from a hidden topic structure of a document and the topic structure is inferred from observable word sequences of document collections. In this study, we allow two distinct word generation to incorporate semantic dependencies between words, particularly between expansions (long forms) of abbreviations and their sentential co-occurring words. Besides topic information, the semantic dependency between words is defined as a link and a new random parameter for the link presence is assigned to each word. As a result, the most probable expansions with respect to abbreviations of a given abstract are decided by word-topic distribution, document-topic distribution, and word-link distribution estimated from document collection though the semantic dependency link topic model. The abstracts retrieved from the MEDLINE Entrez interface by the query relating 22 abbreviations and their 186 expansions were used as a data set. The link topic model correctly predicted expansions of abbreviations with the accuracy of 98.30%.

• The Journal of Information Systems
• /
• v.26 no.4
• /
• pp.327-353
• /
• 2017

Purpose Social media such as social network services, online forums, and customer reviews have produced a plethora amount of information online. Yet, the information deluge has created both opportunities and challenges at the same time. This research particularly focuses on the challenges in order to discover and track the service defects over time derived by mining publicly available online customer reviews. Design/methodology/approach Synthesizing the streams of research from text analytics, we apply two stages of methods of sentiment analysis and structural topic model incorporating meta-information buried in review texts into the topics. Findings As a result, our study reveals that the research framework effectively leverages textual information to detect, prioritize, and categorize service defects by considering the moving trend over time. Our approach also highlights several implications theoretically and practically of how methods in computational linguistics can offer enriched insights by leveraging the online medium.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.8
• /
• pp.3168-3186
• /
• 2020

Automatic topic essay generation (TEG) is a controllable text generation task that aims to generate informative, diverse, and topic-consistent essays based on multiple topics. To make the generated essays of high quality, a reasonable method should consider both diversity and topic-consistency. Another essential issue is the intrinsic link of the topics, which contributes to making the essays closely surround the semantics of provided topics. However, it remains challenging for TEG to fill the semantic gap between source topic words and target output, and a more powerful model is needed to capture the semantics of given topics. To this end, we propose a pretraining-based contextual self-attention (PC-SAN) model that is built upon the seq2seq framework. For the encoder of our model, we employ a dynamic weight sum of layers from BERT to fully utilize the semantics of topics, which is of great help to fill the gap and improve the quality of the generated essays. In the decoding phase, we also transform the target-side contextual history information into the query layers to alleviate the lack of context in typical self-attention networks (SANs). Experimental results on large-scale paragraph-level Chinese corpora verify that our model is capable of generating diverse, topic-consistent text and essentially makes improvements as compare to strong baselines. Furthermore, extensive analysis validates the effectiveness of contextual embeddings from BERT and contextual history information in SANs.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.20 no.5
• /
• pp.207-214
• /
• 2020

Nowadays, the use of smart phones and various electronic devices is increasing, the Internet and SNS are activated, and we live in the flood of information. The amount of information has grown exponentially, making it difficult to look at a lot of information, and more and more people want to see only key keywords in a document, and the importance of research to extract topics that are the core of information is increasing. In addition, it is also an important issue to extract the topic and compare it with the past to infer the current trend. Topic modeling techniques can be used to extract topics from a large volume of documents, and these extracted topics can be used in various fields such as trend prediction and data analysis. In this paper, we inquire the topic of the three-year papers of 2016, 2017, and 2018 in the field of computing using the LDA algorithm, one of Probabilistic Topic Model Techniques, in order to analyze the rapidly changing trends and keep pace with the times. Then we analyze trends and flows of research.

• Journal of Information Processing Systems
• /
• v.17 no.6
• /
• pp.1057-1070
• /
• 2021

Case-related news filtering is crucial in legal text mining and divides news into case-related and case-unrelated categories. Because case-related news originates from various fields and has different writing styles, it is difficult to establish complete filtering rules or keywords for data collection. In addition, the labeled corpus for case-related news is sparse; therefore, to train a high-performance classification model, it is necessary to annotate the corpus. To address this challenge, we propose topic-enhanced positive-unlabeled learning, which selects positive and negative samples guided by topics. Specifically, a topic model based on a variational autoencoder (VAE) is trained to extract topics from unlabeled samples. By using these topics in the iterative process of positive-unlabeled (PU) learning, the accuracy of identifying case-related news can be improved. From the experimental results, it can be observed that the F1 value of our method on the test set is 1.8% higher than that of the PU learning baseline model. In addition, our method is more robust with low initial samples and high iterations, and compared with advanced PU learning baselines such as nnPU and I-PU, we obtain a 1.1% higher F1 value, which indicates that our method can effectively identify case-related news.

• Proceedings of the National Institute of Ecology of the Republic of Korea
• /
• v.4 no.1
• /
• pp.43-48
• /
• 2023

This study analyzed research trends in the field of ecological research. Data were collected based on a keyword search of the SCI, SSCI, and A&HCI databases from January 2002 to September 2022. The seven keywords, including biodiversity, ecology, ecotourism, species, climate change, ecosystem, restoration, wildlife, were recommended by ecological research experts. Word clouds were created for each of the searched keywords, and topic map analysis was performed. Topic map analysis using biodiversity, climate change, ecology, ecosystem, and restoration each generated 10 topics; topic maps analysis using the ecotourism keyword generated 5 topics; and topic map analysis using the wildlife keyword generated 4 topics. Each topic contained six keywords.

• 한국벤처창업학회:학술대회논문집
• /
• 2022.04a
• /
• pp.123-126
• /
• 2022

Online reviews offer valuable information to businesses by reflecting consumer experiences about their products and services. Two important aspects of online reviews are first, the topics consumers choose to address and second, the sentiments expressed in their reviews. Building upon previous literature that shows online reviews are context-dependent, we examine the impact of topic distribution on review sentiment in South Korea and the U.S. during pre-and post-pandemic periods. After performing topic modeling on Airbnb app review data, we measure the contribution of each topic on review sentiment using SHAP values. Our results indicate variations in topic distribution trends between 2018 and 2021. Also, the order and magnitude of topics' impact on review sentiment change between pre-and post-pandemic periods for both countries. This study can help businesses to understand how topics and sentiments associated with their products and services changed after pandemic, and also help them identify areas of improvement.

• Informatization Policy
• /
• v.30 no.1
• /
• pp.41-66
• /
• 2023

This study aims to reveal the direction of municipal ordinances for smart cities, while focusing on 74 municipal ordinances from 72 municipal governments through topic modeling. As a result, the main keywords that show a high frequency belong to establishment and operations of the Smart City Committee. From the result of topic modeling Latent Dirichlet Allocation(LDA), it classifies municipal ordinances for smart cities into eight topics as follows: Topic 1(security for process of smart cities), Topic 2(promotion of smart city industry), Topic 3(composition of a smart city consultative body for local residents), Topic 4(support system for smart cities), Topic 5(management for personal information), Topic 6(use of smart city data), Topic 7(implementation for intelligent public administration), and Topic 8(smart city promotion). As for topic categorization by region, Topics 5, 6, and 8 which are mostly related to the practical operation of smart cities have a significant portion of municipal ordinances for smart cities in the Seoul metropolitan area. Then, Topics 2, 3, and 4 which are mostly related to the initial implementation of smart cities have a significant portion of municipal ordinances for smart cities in provincial areas.

• Journal of the Korean Society for information Management
• /
• v.32 no.1
• /
• pp.153-169
• /
• 2015

The combined approach of using ego-centric network analysis and dynamic citation network analysis for refining the result of LDA-based topic modeling was suggested and examined in this study. Tow datasets were constructed by collecting Web of Science bibliographic records of White LED and topic modeling was performed by setting a different number of topics on each dataset. The multi-assigned top keywords of each topic were re-assigned to one specific topic by applying an ego-centric network analysis algorithm. It was found that the topical cohesion of the result of topic modeling with the number of topic corresponding to the lowest value of perplexity to the dataset extracted by SPLC network analysis was the strongest with the best values of internal clustering evaluation indices. Furthermore, it demonstrates the possibility of developing the suggested approach as a method of multi-faceted research trend detection.