Topic Modeling Techniques Research Articles

A Dynamic Perspective on Publics and Counterpublics: The Role of the Blogosphere in Pushing the Issue of Climate Change During the 2016 US Presidential Campaign

ABSTRACT Climate change was hardly debated during the 2016 US presidential campaign. Against this background and building upon Fraser's concept of counterpublics (1990), this paper examines whether climate change advocates used the English-speaking blogosphere to push their positions forward. This study uses blog data starting from the Republican nomination of Donald Trump (20 July 2016) to Election Day (8 November 2016) and applies a computerized classification algorithm and topic-modeling techniques to explore, first, the salience of skeptic and advocate positions toward climate change in the English-speaking blogosphere and, second, with which topics these positions are most connected. The results show that the positions and topics of climate change advocates were more salient online than those of climate-change skeptics during the 2016 US presidential campaign. Thus, the study shows that the relation between different publics in societal discourses is not static but may change dynamically over time.

Review rating prediction using combined latent topics and associated sentiments: an empirical review

Understanding the tastes of each user and the characteristics of each product is necessary to predict how a user will respond to a new product. These latent user and product dimensions can be discovered with the help of user feedback. A numeric rating and its accompanying text review is the most widely available form of user feedback. A measure which encapsulates the contents of such reviews is often necessary as they have been found to significantly influence the shopping behavior of users. A fine-grained form of such measure that could act as a perfect feedback about the product is star rating. The review rating prediction tries to predict a rating corresponding to the given review. An approach that performs review rating prediction task by using the latent topics extracted from reviews and their associated sentiments is proposed in this paper. The proposed approach treats review rating prediction problem as a multi-class classification problem. An empirical review of topic modeling techniques such as (i) term frequency-inverse document frequency(TF-IDF), (ii) latent Dirichlet allocation(LDA), and (iii) nonnegative matrix factorization(NNMF) is performed in this work to investigate their efficiency. The proposed approach has a lot of advantages: Firstly, this approach accurately predicts product ratings by making use of the topics and their sentiments present in the reviews; this is useful in occasions where only reviews are available. Secondly, the discovered topics and their sentiments can be used to recognize informative reviews. Thirdly, it facilitates to justify the rating with review text. The experimental results show that the proposed model works well with latent topics and sentiments when the underlying model is trained using a deep learning technique.

A survey of Emerging Techniques in Detecting SMS Spam

In the past years, spammers have focused their attention on sending spam through short messages services (SMS) to mobile users. They have had some success because of the lack of appropriate tools to deal with this issue. This paper is dedicated to review and study the relative strengths of various emerging technologies to detect spam messages sent to mobile devices. Machine Learning methods and topic modelling techniques have been remarkably effective in classifying spam SMS. Detecting SMS spam suffers from a lack of the availability of SMS dataset and a few numbers of features in SMS. Various features extracted and dataset used by the researchers with some related issues also discussed. The most important measurements used by the researchers to evaluate the performance of these techniques were based on their recall, precision, accuracies and CAP Curve. In this review, the performance achieved by machine learning algorithms was compared, and we found that Naive Bayes and SVM produce effective performance.

Clustering of Translation via Topic Modeling

In the present study, the authors investigated structural differences between scientific articles arising from their translation from Russian to English. In the course of the research, a modal topic modelling technique was used. Each document in the assembled collection was presented in two versions: English and Russian. As a result of constructing the topic model, the bimodal matrices Φ and Θ were obtained. An analysis of the Φ matrix showed that the topics could be distinguished according to the degree of correspondence between Russian and English terms when considering words in decreasing order of probability. For 90% of the topics, the English words fully corresponded to the Russian words used. Analysis of the Θ matrix showed that for 99% of the documents a topic exists having a value greater than 0.95. Thus, the majority of documents are monotopical. Moreover, this majority does not depend on the language of the document. Although the de facto language of scientific articles nowadays is English, a considerable number of scientific works are initially published in scientists’ respective native languages and only subsequently translated into English in a more complete and in-depth form. Thus, it is possible to speak about the existence of a bilingual corpus of such documents.

Spam Diffusion in Social Networking Media using Latent Dirichlet Allocation

Like web spam has been a major threat to almost every aspect of the current World Wide Web, similarly social spam especially in information diffusion has led a serious threat to the utilities of online social media. To combat this challenge the significance and impact of such entities and content should be analyzed critically. In order to address this issue, this work usedTwitter as a case study and modeled the contents of information through topic modeling and coupled it with the user oriented feature to deal it with a good accuracy. Latent Dirichlet Allocation (LDA) a widely used topic modeling technique is applied to capture the latent topics from the tweets’ documents. The major contribution of this work is twofold: constructing the dataset which serves as the ground-truth for analyzing the diffusion dynamics of spam/non-spam information and analyzing the effects of topics over the diffusibility. Exhaustive experiments clearly reveal the variation in topics shared by the spam and non-spam tweets. The rise in popularity of online social networks, not only attracts legitimate users but also the spammers. Legitimate users use the services of OSNs for a good purpose i.e., maintaining the relations with friends/colleagues, sharing the information of interest, increasing the reach of their business through advertisings.

Tracking biomedicalization in the media: Public discourses on health and medicine in the UK and Italy, 1984–2017

This article examines historical trends in the reporting of health, illness and medicine in UK and Italian newspapers from 1984 to 2017. It focuses on the increasing “biomedicalization” of health reporting and the framing of health and medicine as a matter of technoscientific interventions. Methodologically, we relied on two large datasets consisting of all the health- and medicine-related articles published in the online archives of The Guardian (UK) and la Repubblica (Italy). These articles underwent a quantitative analysis, based on topic modelling techniques, to identify and analyse relevant topics in the datasets. Moreover, we developed some synthetic indices to support the analysis of how medical and health news are “biomedicalized” in media coverage. Theoretically, we emphasise that media represent a constitutive environment in shaping biomedicalization processes. Our analyses show that across the period under scrutiny, biomedicalization is a relevant, even if sometimes ambivalent, frame in the media sphere, placing growing centrality on three dimensions: i) health and well-being as a matter of individual commitment to self-monitoring and self-surveillance; ii) biomedicine as a large technoscientific enterprise emerging from the entanglement between research fields and their technological embodiments; iii) the multiverse reforms of welfare systems in facing the trade-off between universal health coverage and the need to render the national healthcare system more sustainable and compatible with non-expansionary monetary policies and austerity approaches in managing state government budgets.

A Machine Learning Approach to Customer Needs Analysis for Product Ecosystems

AbstractCreating product ecosystems has been one of the strategic ways to enhance user experience and business advantages. Among many, customer needs analysis for product ecosystems is one of the most challenging tasks in creating a successful product ecosystem from both the perspectives of marketing research and product development. In this paper, we propose a machine-learning approach to customer needs analysis for product ecosystems by examining a large amount of online user-generated product reviews within a product ecosystem. First, we filtered out uninformative reviews from the informative reviews using a fastText technique. Then, we extract a variety of topics with regard to customer needs using a topic modeling technique named latent Dirichlet allocation. In addition, we applied a rule-based sentiment analysis method to predict not only the sentiment of the reviews but also their sentiment intensity values. Finally, we categorized customer needs related to different topics extracted using an analytic Kano model based on the dissatisfaction-satisfaction pair from the sentiment analysis. A case example of the Amazon product ecosystem was used to illustrate the potential and feasibility of the proposed method.

Visual topic models for healthcare data clustering

Social media is a great source to search health-related topics for envisages solutions towards healthcare. Topic models originated from Natural Language Processing that is receiving much attention in healthcare areas because of interpretability and its decision making, which motivated us to develop visual topic models. Topic models are used for the extraction of health topics for analyzing discriminative and coherent latent features of tweet documents in healthcare applications. Discovering the number of topics in topic models is an important issue. Sometimes, users enable an incorrect number of topics in traditional topic models, which leads to poor results in health data clustering. In such cases, proper visualizations are essential to extract information for identifying cluster trends. To aid in the visualization of topic clouds and health tendencies in the document collection, we present hybrid topic modeling techniques by integrating traditional topic models with visualization procedures. We believe proposed visual topic models viz., Visual Non-Negative Matrix Factorization (VNMF), Visual Latent Dirichlet Allocation (VLDA), Visual intJNon-negative Matrix Factorization (VintJNMF), and Visual Probabilistic Latent Schematic Indexing (VPLSI) are promising methods for extracting tendency of health topics from various sources in healthcare data clustering. Standard and benchmark social health datasets are used in an experimental study to demonstrate the efficiency of proposed models concerning clustering accuracy (CA), Normalized Mutual Information (NMI), precision (P), recall (R), F-Score (F) measures and computational complexities. VNMF visual model performs significantly at an increased rate of 32.4% under cosine based metric in the display of visual clusters and an increased rate of 35–40% in performance measures compared to other visual methods on different number of health topics.

Improved Topic Modeling with Parallel-Supervised LDA

In the modern era of digitalization, our day-to-day life is entirely dependent on digital platform-from raising our voice in social media to online shopping. Our collective knowledge is continued to be accumulated in the form of electronic texts, blogs, news, images, audios, videos and in many more ways and on account of this there is a greater need of analyzing these huge contents to get rid of the difficulties in searching the object, we aim for. Topic modelling is an efficient machine learning techniques for discovering the hidden semantic structure of contents. “Latent Dirichlet Allocation” (LDA) is a generative probabilistic topic modelling, which is the basis of other generative topic modelling techniques. New models are coming up with advanced algorithms in order to improve the topic modelling. Existing models have their own limitation. In case of obtaining more accuracy, the processing time of topic modelling goes high while in consideration of achieving more speed, accuracy gets low. Most of the algorithms implemented earlier cannot perform well in above-mentioned area. In this paper, we would like to introduce parallel-supervised LDA model where supervised Latent Dirichlet Algorithm (sLDA) and parallel Latent Dirichlet Algorithm (pLDA) are applied together to obtain high accurate results with quicker response time.

Identifying Topical Coverages of Curricula using Topic Modeling and Visualization Techniques: A Case of Digital and Data Curation

Digital/data curation curricula have been around for a couple of decades. Currently, several ALA-accredited LIS programs offer digital/data curation courses and certificate programs to address the high demand for professionals with the knowledge and skills to handle digital content and research data in an ever-changing information environment. In this study, we aimed to examine the topical scopes of digital/data curation curricula in the context of the LIS field. We collected 16 syllabi from the digital/data curation courses, as well as textual descriptions of the 11 programs and their core courses offered in the U.S., Canada, and the U.K. The collected data were analyzed using a probabilistic topic modeling technique, Latent Dirichlet Allocation, to identify both common and unique topics. The results are the identification of 20 topics both at the program- and course-levels. Comparison between the program- and course-level topics uncovered a set of unique topics, and a number of common topics. Furthermore, we provide interactive visualizations for digital/data curation programs and courses for further analysis of topical distributions. We believe that our combined approach of a topic modeling and visualizations may provide insight for identifying emerging trends and co-occurrences of topics among digital/data curation curricula in the LIS field.

A Framework Based on K-Means Clustering and Topic Modeling for Analyzing Unstructured Manufacturing Capability Data

Abstract The natural language descriptions of the capabilities of manufacturing companies can be found in multiple locations including company websites, legacy system databases, and ad hoc documents and spreadsheets. To unlock the value of unstructured capability data and learn from it, there is a need for developing advanced quantitative methods supported by machine learning and natural language processing techniques. This research proposes a hybrid unsupervised learning methodology using K-means clustering and topic modeling techniques in order to build clusters of suppliers based on their capabilities, automatically infer topics from the created clusters, and discover nontrivial patterns in manufacturing capability corpora. The capability data is extracted either directly from the website of manufacturing firms or from their profiles in e-sourcing portals and directories. Feature extraction and dimensionality reduction process in this work are supported by N-gram extraction and latent semantic analysis (LSA) methods. The proposed clustering method is validated experimentally based on a dataset composed of 150 capability descriptions collected from web-based sourcing directories such as the Thomas Net directory for manufacturing companies. The results of the experiment show that the proposed method creates supplier cluster with high accuracy. Two example applications of the proposed framework, related to supplier similarity measurement and automated thesaurus creation, are introduced in this paper.

The usage of large data sets in online consumer behaviour: A bibliometric and computational text-mining–driven analysis of previous research

The paper reports the evolution of scientific research on usage of large datasets in online consumer behaviour between 2000 and 2018. Thus, it affords information regarding the evolution of the field in terms of identifying key publications and authors as well as how certain topics have evolved over time. In addition, by utilising topic modelling and text analytic techniques, it is identified certain research themes from the papers within the published articles included in the dataset. This offers a guide to those who want to contribute to the field. In addition, paper contributes to the methodology related to literature surveys and bibliometric analyses by conducted topic modelling to extract the latent topics from the collected literature by utilising Structural Topic Modelling in order to gain more elaborated results.

A Topic Modeling Approach to the Analysis of Issues for Kids Creators

1인 미디어 크리에이터로 참여하는 아이들의 수가 점차 증가하고 있으며, 이에 대한 사회적 관심이 또한 증대되고 있음에도 불구하고 이에 대한 학술적 논의는 매우 미미한 실정이다. 이에, 본 연구에서는 뉴스 데이터 분석을 통하여 ‘키즈 크리에이터’들이 현재 우리 사회에서 어떠한 이슈를 만들어내고 있으며, 우리 사회는 이들을 어떻게 바라보고 있는지를 포착해보고자 하였다. 빅카인즈(BigKinds)를 통하여 2,947개의 뉴스 데이터를 수집 후 중복 데이터를 제거하고, 적합성 검토를 수행하여 517개의 뉴스 데이터를 최종 분석대상으로 선정하였다. 이어 해당 뉴스 데이터에 등장하는 단어들의 속성을 추출하고 이에 담긴 잠재적인 정보를 발굴하는 토픽 모델링을 수행하여 6개의 토픽을 추출하였다.Recently, the number of children participating as one-person media creators is gradually increasing, and social interest is also increasing. However, there is no comprehensive study on social issues related to the kids creators. This study examines how the kids creators are perceived through news data analysis. This study analyzed the related news data by applying the text mining technique. In other words, the collection of news data was carried out through Bigkines. A total of 2,947 news items were gathered, and the redundancy and relevance were reviewed. Finally, 517 news items were included for the LDA, one of topic modeling techniques. Six major topics were extracted.

VASSL: A Visual Analytics Toolkit for Social Spambot Labeling

Social media platforms are filled with social spambots. Detecting these malicious accounts is essential, yet challenging, as they continually evolve to evade detection techniques. In this article, we present VASSL, a visual analytics system that assists in the process of detecting and labeling spambots. Our tool enhances the performance and scalability of manual labeling by providing multiple connected views and utilizing dimensionality reduction, sentiment analysis and topic modeling, enabling insights for the identification of spambots. The system allows users to select and analyze groups of accounts in an interactive manner, which enables the detection of spambots that may not be identified when examined individually. We present a user study to objectively evaluate the performance of VASSL users, as well as capturing subjective opinions about the usefulness and the ease of use of the tool.

Online Interactive Collaborative Filtering Using Multi-Armed Bandit with Dependent Arms

Online interactive recommender systems strive to promptly suggest to consumers appropriate items (e.g., movies, news articles) according to the current context including both the consumer and item content information. However, such context information is often unavailable in practice for the recommendation, where only the users' interaction data on items can be utilized. Moreover, the lack of interaction records, especially for new users and items, worsens the performance of recommendation further. To address these issues, collaborative filtering (CF), one of the recommendation techniques relying on the interaction data only, as well as the online multi-armed bandit mechanisms, capable of achieving the balance between exploitation and exploration, are adopted in the online interactive recommendation settings, by assuming independent items (i.e., arms). Nonetheless, the assumption rarely holds in reality, since the real-world items tend to be correlated with each other (e.g., two articles with similar topics). In this paper, we study online interactive collaborative filtering problems by considering the dependencies among items. We explicitly formulate the item dependencies as the clusters on arms, where the arms within a single cluster share the similar latent topics. In light of the topic modeling techniques, we come up with a generative model to generate the items from their underlying topics. Furthermore, an efficient online algorithm based on particle learning is developed for inferring both latent parameters and states of our model. Additionally, our inferred model can be naturally integrated with existing multi-armed selection strategies in the online interactive collaborating setting. Empirical studies on two real-world applications, online recommendations of movies and news, demonstrate both the effectiveness and efficiency of the proposed approach.

토픽 모델링을 통한 공유숙박업의 이슈 분석: 뉴스 데이터를 중심으로

Recently, accommodation sharing service has been getting attention as a new field of starting a business because of the low entry barriers for business. The government recognized the sharing service as a new opportunity for economic development and the Korean government pushed ahead with the policy for promoting the accommodation sharing service. It has been six years since the introduction of accommodation sharing service in Korea. However, there is no comprehensive study on social issues related to accommodation sharing service. This study examines how accommodation sharing service is perceived as new industry through news data analysis. This study analyzed the related news data by applying the text mining technique. In other words, the collection of news data was carried out through Bigkines. A total of 358 news items were gathered, and the redundancy and relevance were reviewed. Finally, 172 news items were included for the LDA, one of topic modeling techniques. Five major topics were extracted.

Topic Modeling the Research‐Practice Gap in Public Administration

Abstract The possible existence of a research‐practice gap is the topic of a long‐standing debate in the field of public administration. In this Viewpoint essay, the authors examine the agendas of scholars and practitioners using the topic modeling technique of computational social science. Topic modeling a content analysis of 35 topics identified in Public Administration Review and PA Times (3,796 articles) showed that just over 50 percent of topics were common to both groups, indicating shared interests. There were, however, topics that the two groups distinctly focused on. Moreover, scholars and practitioners attached significant differences to the weights allocated to the prominent topics in their writing. Taken together, these findings indicate that topic modeling can shed new light on the research‐practice gap in public administration.

Seven Approaches to Manage Complex Coupled Human and Natural Systems: A Sustainability Toolbox.

Since the publication of the Report of the World Commission on Environment and Development in 1987, there have been numerous studies on sustainability. These studies created new knowledge and tools for understanding and managing complex coupled human and natural systems. In this Critical Review, we used a topic modeling technique to analyze 12 526 peer-reviewed research articles and identify the research questions and the approaches that were used or developed in each of the studies. These approaches were then classified by function. The analysis revealed twenty-three categories of research questions and seven functional approach classes-design for sustainability, modeling of complexity, sustainability indicators, life cycle sustainability assessment, decision making support, sustainability governance, and engagement-each of which is described here as an individual approach or tool within a larger sustainability toolbox. The article concludes with a discussion about using the sustainability toolbox as an integrated knowledge system to support transdisciplinary study and decision-making.

Analysis of launch strategy in cross-border e-Commerce market via topic modeling of consumer reviews

Spurred by the policy of China’s Belt and Road Initiative, Chinese e-Commerce companies have found great opportunities in selling goods overseas. The cross-border e-Commerce shares similarities of launch and marketing strategies with domestic e-Commerce, but also has substantial differences. How to make strategic adjustments to better adapt to the overseas market is of great concern to cross-border e-Commerce companies. Analyzing behaviors of overseas consumers could offer an effective way to address this issue and has attracted great interest of researchers. Consumer comments, cheap and abundant by its nature, provides an easy access for analysis of consumer behaviors. In this paper, we focus on consumer reviews of a specific product, the cellphones, and apply topic modeling techniques to investigate the differences between behaviors of domestic and overseas consumers. We find that consumers from domestic and overseas focus on different aspects of product. In addition, the foreign consumers care more about product quality and tend to make description of technique details. On the contrary, domestic buyers pay more attention on consumer services and intend to comment in generalities. All these findings could help e-Commerce companies design better launch strategies in cross-border e-Commerce market.

Exploring Spatiotemporal Characteristics of Twitter data Using Topic Modelling Techniques

Abstract. This research purpose aims to explore the spatiotemporal aspects of social media data with Twitter data by using topic modelling techniques. The spatiotemporal limits are restricted in two areas of the Republic of Korea: Seoul and Jeju Island. This paper searches the semantics and geographical place characteristics of the Twitter data, and the semantics and place characteristics are regarded as topics in the topic modelling. This paper also discusses the temporal intensity over different spatial areas and visualizes the spatiotemporal patterns with GIS techniques.As Twitter mobility message contains a user’s interests and behavioural patterns in the geo-tagged data corresponding to its location, it is possible to explore geographical locality and the user’s mobility over space by using textual ontological techniques such as topic modelling. Therefore, this paper attempts keywords searching and textural classification to classify the shared spatial activity patterns of the Twitter users. Consequently, two main analysis themes are explored: the tourist activity patterns attracting the visitors in Jeju over time and temporal periodicity for shopping and meal preference in Seoul, respectively.In conclusion, this research represents a potential of the social network data that enables to fill the gap of spatiotemporal patterns of human beings over the online and mobile environment. Furthermore, our study confirms social data analysis techniques as an alternative geographical data source that can complement and replace the roles of spatial data, which could not be analysed in the conventional offline data.