Search

Political parties are under increasing pressure to extend and activate their voter bases by employing more innovative communication strategies. This article focuses on the social media platform Twitter to explore how well Swiss parties performed in terms of employing digital communication during the 2015 federal election campaign. As such, it uses the follower network as an indicator of organizational cohesion, along with two indicators of programmatic coherence based on Twitter message content. Computing density and centrality statistics allow for the quantification of these two aspects in the party networks, while the nonparametric bootstrap introduces uncertainty of the account sampling process into the analysis. Our results suggest that smaller and newer parties, as well as the Social Democrats, tend to exhibit disproportionally high levels of organizational cohesion. At the same time, most parties show comparable—and also disproportionately low—levels of programmatic coherence compared to those displayed by the Social Democrats. [ABSTRACT FROM AUTHOR]

Latent Dirichlet allocation model (LDA) has been widely used in topic modeling. Recent works have shown the effectiveness of integrating neural network mechanisms with this generative model for learning text representation. However, one of the significant setbacks of LDA is that it is based on a Dirichlet prior that has a restrictive covariance structure. All its variables are considered to be negatively correlated, which makes the model restrictive. In a practical sense, topics can be positively or negatively correlated. To address this problem, we proposed a generalized Dirichlet variational autoencoder (GD-VAE) for topic modeling. The Generalized Dirichlet (GD) distribution has a more general covariance structure than the Dirichlet distribution because it takes into account both positively and negatively correlated topics in the corpus. Our proposed model leverages rejection sampling variational inference using a reparameterization trick for effective training. GD-VAE compares favorably to recent works on topic models on several benchmark corpora. Experiments show that accounting for topics' positive and negative correlations results in better performance. We further validate the superiority of our proposed framework on two image data sets. GD-VAE demonstrates its significance as an integral part of a classification architecture. For reproducibility and further research purposes, code for this work can be found at https://github.com/hormone03/GD-VAE. • We propose GD-VAE to capture correlations and learn complex distributions. • We show that capturing all correlations leads to improved performance in GD-VAE. • We address training instability by introducing a weighted objective function. • The comprehensive experiments show that GD-VAE outperformed state-of-the-art models. • We demonstrate the effectiveness of GD-VAE on data augmentation with image data sets. [ABSTRACT FROM AUTHOR]

This work considers a computationally and statistically efficient parameter estimation method for a wide class of latent variable models-including Gaussian mixture models, hidden Markov models, and latent Dirichlet allocation-which exploits a certain tensor structure in their low-order observable moments (typically, of second- and third-order). Specifically, parameter estimation is reduced to the problem of extracting a certain (orthogonal) decomposition of a symmetric tensor derived from the moments; this decomposition can be viewed as a natural generalization of the singular value decomposition for matrices. Although tensor decompositions are generally intractable to compute, the decomposition of these specially structured tensors can be efficiently obtained by a variety of approaches, including power iterations and maximization approaches (similar to the case of matrices). A detailed analysis of a robust tensor power method is provided, establishing an analogue of Wedin's perturbation theorem for the singular vectors of matrices. This implies a robust and computationally tractable estimation approach for several popular latent variable models.

This work considers a computationally and statistically efficient parameter estimation method for a wide class of latent variable models-including Gaussian mixture models, hidden Markov models, and latent Dirichlet allocation-which exploits a certain tensor structure in their low-order observable moments (typically, of second- and third-order). Specifically, parameter estimation is reduced to the problem of extracting a certain (orthogonal) decomposition of a symmetric tensor derived from the moments; this decomposition can be viewed as a natural generalization of the singular value decomposition for matrices. Although tensor decompositions are generally intractable to compute, the decomposition of these specially structured tensors can be efficiently obtained by a variety of approaches, including power iterations and maximization approaches (similar to the case of matrices). A detailed analysis of a robust tensor power method is provided, establishing an analogue of Wedin's perturbation theorem for the singular vectors of matrices. This implies a robust and computationally tractable estimation approach for several popular latent variable models.

Abstract The study of biodiversity has grown exponentially in the last thirty years in response to demands for greater understanding of the function and importance of Earth's biodiversity and finding solutions to conserve it. Here, we test the hypothesis that biodiversity science has become more interdisciplinary over time. To do so, we analyze 97,945 peer‐reviewed articles over a twenty‐two‐year time period (1990–2012) with a continuous time dynamic model, which classifies articles into concepts (i.e., topics and ideas) based on word co‐occurrences. Using the model output, we then quantify different aspects of interdisciplinarity: concept diversity, that is, the diversity of topics and ideas across subdisciplines in biodiversity science, subdiscipline diversity, that is, the diversity of subdisciplines across concepts, and network structure, which captures interactions between concepts and subdisciplines. We found that, on average, concept and subdiscipline diversity in biodiversity science were either stable or declining, patterns which were driven by the persistence of rare concepts and subdisciplines and a decline in the diversity of common concepts and subdisciplines, respectively. Moreover, our results provide evidence that conceptual homogenization, that is, decreases in temporal β concept diversity, underlies the observed trends in interdisciplinarity. Together, our results reveal that biodiversity science is undergoing a dynamic phase as a scientific discipline that is consolidating around a core set of concepts. Our results suggest that progress toward addressing the biodiversity crisis via greater interdisciplinarity during the study period may have been slowed by extrinsic factors, such as the failure to invest in research spanning across concepts and disciplines. However, recent initiatives such as the Intergovernmental Science‐Policy Platform on Biodiversity and Ecosystem Services (IPBES) may attract broader support for biodiversity‐related issues and hence interdisciplinary approaches to address scientific, political, and societal challenges in the coming years.

We provide a brief, non-technical introduction to the text mining methodology known as topic modeling. We summarize the theory and background of the method and discuss just what kinds of things are found by topic models. Using a text corpus comprised of the eight articles from the special issue of Poetics on the subject of topic models, we run a topic model on these papers both as a way to introduce the methodology and also to help summarize some of the ways in which social and cultural scientists are using topic models.  We review some of the critiques and debates over the use of the method and finally, we link these developments back to some of the original innovations in the field of content analysis that were pioneered by Harold D. Lasswell and colleagues during and just after World War II.

The aim of the article is to detect and review existing topic modelling methods of Latent Dirichlet Allocation and their modifications usable in Voice of Customer analysis. Voice of Customer is expressed mainly through textual comments which often focus on the evaluation of products or services the customer consumes. The most studied data source are customer reviews which contain next to the textual comments also ratings in form of scales. The aim of the topic models is to mine the topics and their aspects the customers are evaluating in their reviews and assign to them a particular sentiment or emotion. The author completed a systematic literature review of peer-reviewed published journal articles indexed in leading databases of Scopus and Web of Science and concerning the current use of Latent Dirichlet Allocation model variants in Voice of Customer textual analysis for performing the tasks of aspect detection, emotion detection, personality detection and sentiment assignation. In total, 38 modifications of the LDA model were identified with the reference to their first application in the research of text analytics. The review is intended for researchers in customer analytics the field of sentiment or emotion detection, and moreover as results from the review, for studies in personality recognition based on the textual data. The review offers a basic overview and comparison of LDA modifications which can be considered as a knowledge baseline for selection in a specific application. The scope of the literature examination is limited to the period of years 2003-2018 with the application relevant to the analysis of Voice of Customer subjective textual data only which is closely connected to the area of marketing or customer relationship management.

The popularity of research topics in clinical psychology has always been changing over time. In this study, we use Latent Dirichlet Allocation (LDA), a well-established statistical modeling approach in machine learning, to extract hot research topics in published review articles in clinical psychology. In Study 1, we use LDA to extract existing topics between 1981 to 2018 from the review articles published on three premium journals in clinical psychology. Results provide stable information about all topics and their proportions. In Study 2, we use a dynamic variant of LDA to identify the development of hot topics from 2007 to 2018. Results show that meta-analysis, psychotherapy, professional development, and depression constantly stay as hot topics all over the 12 years. We also find that behavior intervention has a clear rising trend since 2007. Our results provide a comprehensive summary of the popularity of research topics in clinical psychology in the last couple of years, and the results here can help clinical researchers form a structured view of past research and plan future research directions. [ABSTRACT FROM AUTHOR]

In this article, we introduce a dynamic generative model, the Bayesian allocation model (BAM), for modeling count data. BAM covers various probabilistic nonnegative tensor factorization (NTF) and topic models under one general framework. In BAM, allocations are made using a Bayesian network, whose conditional probability tables can be integrated out analytically. We show that, when allocations are viewed as sequential, the resulting marginal process is a special type of Pólya urn process, which we name as Pólya-Bayes process, an integer valued self-reinforcing process. Exploiting the Pólya urn construction, we develop a novel sequential Monte Carlo (SMC) algorithm for marginal likelihood estimation in BAM, leading to a unified scoring method for discrete variable Bayesian networks with hidden nodes, including various NTF and topic models. The SMC estimator for marginal likelihood has the remarkable property of being unbiased in contrast to variational algorithms which are generally biased. We also demonstrate how our novel SMC-based likelihood estimation can be integrated within a Markov chain Monte Carlo algorithm for a principled and correct (in terms of respecting the true posterior distribution) Bayesian model selection and hyperparameter estimation for BAM. We provide several numerical examples, both on artificial and real datasets, that demonstrate the performance of the algorithms for various data regimes. [ABSTRACT FROM AUTHOR]

We compare several unsupervised probabilistic machine learning methods for market basket analysis, namely binary factor analysis, two topic models (latent Dirichlet allocation and the correlated topic model), the restricted Boltzmann machine and the deep belief net. After an overview of previous applications of unsupervised probabilistic machine learning methods to market basket analysis we shortly present the methods which we investigate and outline their estimation. Performance is measured by tenfold cross-validated log likelihood values. Binary factor analysis vastly outperforms topic models. The restricted Boltzmann machine attains a similar performance advantage over binary factor analysis. Overall, a deep belief net with 45 variables in the first and 15 variables in the second hidden layers turns out to be the best model. We also compare the investigated machine learning methods with respect to ease of interpretation and runtimes. In addition, we show how to interpret the relationships between hidden variables and observed category purchases. To demonstrate managerial implications we estimate the effect of promoting each category both on purchase probability increases of other product categories and the relative increase of basket size. Finally, we indicate several possibilities to extend restricted Boltzmann machines and deep belief nets for market basket analysis. [ABSTRACT FROM AUTHOR]

e-mail scams have been around for a long time, and they are still remarkably successful. Understanding these scams can help combat fraud by raising public awareness and improving anti-fraud technologies. We conducted an in-depth analysis of e-mail contents with a mixed-method approach using machine-learning-based content analysis and qualitative analysis. We identified five stories (i.e., topics) scammers used in their e-mails. Further, we explored similarities between messages from the same scammers. Our findings suggest some scammers resend a message with the same story as before but with highly textual differences. We argue this might be a deceptive maneuver to avoid spam blockers. [ABSTRACT FROM AUTHOR]

The concept of Earth system science denotes a shift in the scientific discourse from disciplinary accounts of isolated components of the global environment towards the holistic and interdisciplinary treatment of their complex, functional interactions. We measure to what extent the environmental scientific literature of the past three decades reflects this system turn. Our initial dataset consists of 133,670 articles published in 95 relevant journals since 1990. We apply a combination of topic modelling and network analysis. Correlated Topic Models identify latent themes ("topics") in the scientific discourse as well as intertopic correlations. This allows the generation of topic networks and thus the application of network-analytic techniques. We generate and analyze 2 topic networks. The first network focuses on climate linkages in a subset of our corpus consisting only of environmental journals without a climate-specific orientation: as the climate system is constitutional for the notion of Earth system science, a system turn should reflect itself in strong linkages between climate- and non-climate environmental topics. The second network, based on the full dataset, applies community detection to identify broader topical clusters. Here, we expect the system turn to manifest itself in communities with high topical diversity regarding the components of the global environment as well as types of human-nature interactions, rather than reflecting the boundaries between broader research fields. Our results show that climate topics are comparatively weakly connected and less integrated into broader thematic packages than other topics; that linkages frequently reflect conceptual debates rather than functional interactions between substantive environmental components; and that the scientific discourse splits into 4 broader domains, two of those being topically homogeneous and the other two comprising only marginally diverse topics. We conclude that the concept of Earth system science is primarily aspirational in nature rather than reflecting an empirical shift in the structure of the scientific discourse. [ABSTRACT FROM AUTHOR]

COVID-19 is having a tremendous impact on gender relations, as care needs have been magnified due to schools and day-care closures. Using topic modeling on over 1,100 open reports from a survey fielded during the first four weeks of the lockdown in Germany, we shed light on how personal experiences of the lockdown differ between women and men. Our results show that, in general, people were most concerned about social contacts and childcare. However, we find clear differences among genders: women worried more about childcare while men were more concerned about paid work and the economy. We argue that the COVID-19 pandemic is affecting women more heavily than men not only at the physical level of work (e.g. women are reducing more paid work hours than men), but also through increasing the division regarding the cognitive level of work (e.g. women are more worried about childcare work while men are about paid work). These developments can potentially contribute to a future widening of the gender wage gap during the recovery process. [ABSTRACT FROM AUTHOR]

User-generated content on numerous sites is indicative of users’ sentiment towards many issues, from daily food intake to using new products. Amid the active usage of social networks and micro-blogs, notably during the COVID-19 pandemic, we may glean insights into any product or service through users’ feedback and opinions. Thus, it is often difficult and time consuming to go through all the reviews and analyse them in order to recognize the notion of the overall goodness or badness of the reviews before making any decision. To overcome this challenge, sentiment analysis has been used as an effective rapid way to automatically gauge consumers’ opinions. Large reviews will possibly encompass both positive and negative opinions on different features of a product/service in the same review. Therefore, this paper proposes an aspect-oriented sentiment classification using a combination of the prior knowledge topic model algorithm (SA-LDA), automatic labelling (SentiWordNet) and ensemble method (Stacking). The framework is evaluated using the dataset from different domains. The results have shown that the proposed SA-LDA outperformed the standard LDA. In addition, the suggested ensemble learning classifier has increased the accuracy of the classifier by more than ~3% when it is compared to baseline classification algorithms. The study concluded that the proposed approach is equally adaptable across multi-domain applications.

Digital media merges disparate places, communities, and topics in reporting about current events, with both local and international impacts. This study maps the semantic spaces of reporting on Catalonia's independence movement in 2015 onto geographic space, forming a "mediascape" of interwoven themes and places. It uses topic modeling to determine semantic themes and geographic information retrieval to map the geographies of those themes. A global pattern of distance decay from Barcelona, the center of the Catalonian independence movement, is apparent. This study also compares the particular media themes of professional sport, which is a globalizing force in Catalonia's geopolitical narrative, and the Scottish independence referendum of 2014. It examines the particular geographies that appear in news articles pertaining to each theme. Contrasting the geographies of Catalonian independence and its relation to soccer and Scottish independence through the media produces a spatial–semantic illustration of how mediascapes form from and influence global geopolitical connections. [ABSTRACT FROM AUTHOR]

The Proceedings of the Ecological Society of New Zealand (PESNZ) and its continuation, the New Zealand Journal of Ecology (NZJE), published more than 1250 articles over the 58 years from 1961-2019. Over this period, the emphasis of ecology as a science and the social context in which it is embedded have changed. Here we provide a bibliometric analysis of the history of the PESNZ and the NZJE to assess how the dominant research themes have changed through time, and the extent to which they reflect broader trends in the policy and funding landscapes. The journals' consistent focus has been on applied ecological issues, especially the effects and control of invasive mammals. However, the most discussed taxa have shifted over time (from deer to brushtail possums to rodents and mustelids). Collaboration has altered dramatically, with few author networks and single-author articles in early issues versus multi-author articles and widespread networks today. Thus, the research published by the Society reflects NZ-specific concerns and broader trends in knowledge production (e.g. the shift to team-based science). We conclude by considering the publications of the NZ Ecological Society through the lens of journals being 'clubs' for the social production of shared knowledge. [ABSTRACT FROM AUTHOR]

Probabilistic topic modeling has been applied in a variety of industrial applications. Training a high-quality model usually requires a massive amount of data to provide comprehensive co-occurrence information for the model to learn. However, industrial data such as medical or financial records are often proprietary or sensitive, which precludes uploading to data centers. Hence, training topic models in industrial scenarios using conventional approaches faces a dilemma: A party (i.e., a company or institute) has to either tolerate data scarcity or sacrifice data privacy. In this article, we propose a framework named Industrial Federated Topic Modeling (iFTM), in which multiple parties collaboratively train a high-quality topic model by simultaneously alleviating data scarcity and maintaining immunity to privacy adversaries. iFTM is inspired by federated learning, supports two representative topic models (i.e., Latent Dirichlet Allocation and SentenceLDA) in industrial applications, and consists of novel techniques such as private Metropolis-Hastings, topic-wise normalization, and heterogeneous model integration. We conduct quantitative evaluations to verify the effectiveness of iFTM and deploy iFTM in two real-life applications to demonstrate its utility. Experimental results verify iFTM's superiority over conventional topic modeling. [ABSTRACT FROM AUTHOR]

The exponential growth of social media has brought with it an increasing propagation of hate speech and hate based propaganda. Hate speech is commonly defined as any communication that disparages a person or a group on the basis of some characteristics such as race, colour, ethnicity, gender, sexual orientation, nationality, religion. Online hate diffusion has now developed into a serious problem and this has led to a number of international initiatives being proposed, aimed at qualifying the problem and developing effective counter-measures. The aim of this paper is to analyse the knowledge structure of hate speech literature and the evolution of related topics. We apply co-word analysis methods to identify different topics treated in the field. The analysed database was downloaded from Scopus, focusing on a number of publications during the last thirty years. Topic and network analyses of literature showed that the main research topics can be divided into three areas: "general debate hate speech versus freedom of expression","hate-speech automatic detection and classification by machine-learning strategies", and "gendered hate speech and cyberbullying". The understanding of how research fronts interact led to stress the relevance of machine learning approaches to correctly assess hatred forms of online speech. [ABSTRACT FROM AUTHOR]

Numerous studies over the past generation have identified germline variants that increase specific cancer risks. Simultaneously, a revolution in sequencing technology has permitted high-throughput annotations of somatic genomes characterizing individual tumors. However, examining the relationship between germline variants and somatic alteration patterns is hugely challenged by the large numbers of variants in a typical tumor, the rarity of most individual variants, and the heterogeneity of tumor somatic fingerprints. In this article, we propose statistical methodology that frames the investigation of germline-somatic relationships in an interpretable manner. The method uses meta-features embodying biological contexts of individual somatic alterations to implicitly group rare mutations. Our team has used this technique previously through a multilevel regression model to diagnose with high accuracy tumor site of origin. Herein, we further leverage topic models from computational linguistics to achieve interpretable lower-dimensional embeddings of the meta-features. We demonstrate how the method can identify distinctive somatic profiles linked to specific germline variants or environmental risk factors. We illustrate the method using The Cancer Genome Atlas whole-exome sequencing data to characterize somatic tumor fingerprints in breast cancer patients with germline BRCA1/2 mutations and in head and neck cancer patients exposed to human papillomavirus., (© 2024 Wiley Periodicals LLC.)

Background: Concerns with problematic research are primarily attributed to statistics and methods used to support data. Language, as an extended component of problematic research in published work, is rarely given the same attention despite language's equally important role in shaping the discussion and framings of presented data.Purpose: This study uses a topic modeling approach to study language as a predictor of potential bias among collected publication histories of several health research areas.Methods: We applied Latent Dirichlet Allocation (LDA) topic models to dissect publication histories disaggregated by three factors commonly cited as language influencers: (1) time, to study ADHD pharmacotherapy; (2) funding source, to study sugar consumption; and (3) nation of origin, to study Pediatric Highly-Active Anti-Retroviral Therapy (P-HAART).Results: We found that, for each factor, there were notable differences in language among each corpus when disaggregated by each factor. For time, article content changed to reflect new trends and research practices for the commonly prescribed ADHD medication, Ritalin. For funding source, industry and federally funded studies had differing foci, despite testing the same hypothesis. For nation of origin, regulatory structures between the United States and Europe seemingly influenced the direction of research.Conclusion: This work presents two contributions to ethics research: (1) language and language framing should be studied as carefully as numeric data among studies of rigor, reproducibility, and transparency; and (2) the scientific community should continue to apply topic models as mediums to answer hypothesis-driven research questions.

Q fever, caused by the zoonotic bacterium Coxiella burnetii, is a globally distributed emerging infectious disease. Livestock are the most important zoonotic transmission sources, yet infection in people without livestock exposure is common. Identifying potential exposure pathways is necessary to design effective interventions and aid outbreak prevention. We used natural language processing and graphical network methods to provide insights into how Q fever notifications are associated with variation in patient occupations or lifestyles. Using an 18‐year time‐series of Q fever notifications in Queensland, Australia, we used topic models to test whether compositions of patient answers to follow‐up exposure questionnaires varied between demographic groups or across geographical areas. To determine heterogeneity in possible zoonotic exposures, we explored patterns of livestock and game animal co‐exposures using Markov Random Fields models. Finally, to identify possible correlates of Q fever case severity, we modelled patient probabilities of being hospitalized as a function of particular exposures. Different demographic groups consistently reported distinct sets of exposure terms and were concentrated in different areas of the state, suggesting the presence of multiple transmission pathways. Macropod exposure was commonly reported among Q fever cases, even when exposure to cattle, sheep or goats was absent. Males, older patients and those that reported macropod exposure were more likely to be hospitalized due to Q fever infection. Our study indicates that follow‐up surveillance combined with text modelling is useful for unravelling exposure pathways in the battle to reduce Q fever incidence and associated morbidity. [ABSTRACT FROM AUTHOR]

Latent Dirichlet Allocation (LDA) is a popular tool for analyzing discrete count data such as text and images, which are required to model datasets and a large number of topics, e.g., tens of thousands of topics for industry scale applications. Although distributed CPU systems have been used to address this problem, they are slow and resource inefficient. GPU-based systems have emerged as a promising alternative because of their high computational power and memory bandwidth. However, existing GPU-based LDA systems can only learn thousands of topics, because they use dense data structures, and have linear time complexity to the number of topics. In this article, we propose SaberLDA, a GPU-based LDA system that implements a sparsity-aware algorithm to achieve sublinear time complexity to learn a large number of topics. To address the challenges introduced by sparsity, we propose a novel data layout, a warp-based sampling kernel, an efficient sparse matrix counting method, and a fine-grained load balancing strategy. SaberLDA achieves linear speedup on 4 GPUs and is 6–10 times faster than existing GPU systems in thousands of topics. It can learn 40,000 topics from a dataset of billions of tokens in two hours, which was previously only achievable using clusters of tens of CPU servers. [ABSTRACT FROM AUTHOR]

Searching for similar documents and exploring major themes covered across groups of documents are common activities when browsing collections of scientific papers. This manual knowledge-intensive task can become less tedious and even lead to unexpected relevant findings if unsupervised algorithms are applied to help researchers. Most text mining algorithms represent documents in a common feature space that abstract them away from the specific sequence of words used in them. Probabilistic Topic Models reduce that feature space by annotating documents with thematic information. Over this low-dimensional latent space some locality-sensitive hashing algorithms have been proposed to perform document similarity search. However, thematic information gets hidden behind hash codes, preventing thematic exploration and limiting the explanatory capability of topics to justify content-based similarities. This paper presents a novel hashing algorithm based on approximate nearest-neighbor techniques that uses hierarchical sets of topics as hash codes. It not only performs efficient similarity searches, but also allows extending those queries with thematic restrictions explaining the similarity score from the most relevant topics. Extensive evaluations on both scientific and industrial text datasets validate the proposed algorithm in terms of accuracy and efficiency. [ABSTRACT FROM AUTHOR]

Analyzing and understanding Internet news are important for many applications, such as market sentiment investigation and crisis management. However, it is challenging for users to interpret a massive amount of unstructured text, to dig out its accurate meaning, and to spot noteworthy news events. To overcome these challenges, we propose a novel visualization-driven approach for analyzing news text. We first collect Internet news from different sources and encode sentences into a vector representation suitable for input to a neural network, which calculates a sentiment score, to help detect news event patterns. A subsequent interactive visualization framework allows the user to explore the development of and relationships between Internet news topics. In addition, a method for detecting news events enables users and domain experts to interactively explore the correlations between market sentiment, topic distribution, and event patterns. We use this framework to provide a web-based interactive visualization system. We demonstrate the applicability and effectiveness of our proposed system using case studies involving blockchain news. [ABSTRACT FROM AUTHOR]

Deep learning architectures based on self-attention have recently achieved and surpassed state of the art results in the task of unsupervised aspect extraction and topic modeling. While models such as neural attention-based aspect extraction (ABAE) have been successfully applied to user-generated texts, they are less coherent when applied to traditional data sources such as news articles and newsgroup documents. In this work, we introduce a simple approach based on sentence filtering in order to improve topical aspects learned from newsgroups-based content without modifying the basic mechanism of ABAE. We train a probabilistic classifier to distinguish between out-of-domain texts (outer dataset) and in-domain texts (target dataset). Then, during data preparation we filter out sentences that have a low probability of being in-domain and train the neural model on the remaining sentences. The positive effect of sentence filtering on topic coherence is demonstrated in comparison to aspect extraction models trained on unfiltered texts. [ABSTRACT FROM AUTHOR]

Recent in situ multiplexed profiling techniques provide insight into microenvironment formation, maintenance, and transformation through a lens of localized cellular phenotype distribution. In this article, we introduce a method for recovering signatures of microenvironments from such data. We use topic models to identify characteristic cell types overrepresented in neighborhoods that serve as proxies for microenvironment. Furthermore, by assuming spatial coherence among neighboring microenvironments our model limits the number of parameters that need to be learned and permits data-driven decisions about the size of cellular neighborhoods. We apply this method to uncover anatomically known structures in mouse spleen—identifying distinct population of spleen B cells that are defined by their characteristic neighborhoods. Next, we apply the method to a dataset of triple-negative breast cancer tumors from 41 patients to study the structure of tumor-immune boundary. We uncover previously reported tumor-immune microenvironment near the tumor-immune boundary enriched for immune cells with high Indoleamine-pyrrole 2,3-dioxygenase (IDO) and Programmed death-ligand 1 (PD-L1) and a novel, immunosuppressed, microenvironment-enriched for cells expressing CD45 and FoxP3. [ABSTRACT FROM AUTHOR]

Topic models have become popular tools for dimension reduction and exploratory analysis of text data which consists in observed frequencies of a vocabulary of p words in n documents, stored in a p x n matrix. The main premise is that the mean of this data matrix can be factorized into a product of two non-negative matrices: a p x K word-topic matrix A and a K x n topic-document matrix W. This paper studies the estimation of A that is possibly element-wise sparse, and the number of topics K is unknown. In this under-explored context, we derive a new minimax lower bound for the estimation of such A and propose a new computationally efficient algorithm for its recovery. We derive a finite sample upper bound for our estimator, and show that it matches the minimax lower bound in many scenarios. Our estimate adapts to the unknown sparsity of A and our analysis is valid for any finite n, p, K and document lengths. Empirical results on both synthetic data and semi-synthetic data show that our proposed estimator is a strong competitor of the existing state-of-the-art algorithms for both non-sparse A and sparse A, and has superior performance is many scenarios of interest. [ABSTRACT FROM AUTHOR]

We develop a topic model with multiple latent changepoints and demonstrate an approach to detect changes in the topics mentioned in brand-related social media posts. Although social media has emerged as a popular source of insights for both researchers and practitioners, much of the work on the dynamics in social media has focused on common metrics such as volume and sentiment. In this research, we develop a changepoint model to capture the underlying shifts in social media content. We extend latent Dirichlet allocation (LDA), a topic modeling approach, by incorporating multiple latent changepoints through a Dirichlet process hidden Markov model that allows for the prevalence of topics to differ before and after each changepoint without requiring prior knowledge about the number of changepoints. We demonstrate our modeling framework using social media posts from brand crises (Volkswagen's 2015 emissions testing scandal and Under Armour's 2018 data breach) and a new product launch (Burger King's 2016 launch of the Angriest Whopper). We show that our model identifies shifts in the conversation surrounding each of these events and outperforms both static and other dynamic topic models. We demonstrate how the model may be used by marketers to actively monitor conversations around their brands, including distinguishing between changes in the conversation arising from a shift in the contributor base and underlying changes in the topics discussed by contributors. [ABSTRACT FROM AUTHOR]

Computational methods offer a new perspective on the evolving agendas of right-wing movements and parties online. This article showcases computational approaches to text analysis (specifically so-called topic models) to diachronically investigate nativist right-wing issues in social media by comparing comments posted on the Facebook page of the Pegida movement to those of the Alternative for Germany. After describing topic modelling as an increasingly popular method and drawing on the literature on right-wing nativism online, we investigate a set of shared issues relevant to the mobilization of commentators, including opposition to Islam, migration, the government and the media. We furthermore show contrastively how issue prevalence differs between the two groups, and how issue shares change over time, in some instances converging on a shared nativist core. We close with a series of suggestions on the utility of computation content analysis for the study of rapidly evolving political agendas. [ABSTRACT FROM AUTHOR]

Copyright of European Policy Analysis is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Renewed efforts at empirically distinguishing between different forms of political regimes leave out the cultural dimension. In this article, we demonstrate how modern computational tools can be used to fill this gap. We employ web-scraping techniques to generate a data set of speeches by heads of government in European democracies and autocratic regimes around the globe. Our data set includes 4740 speeches delivered between 1999 and 2019 by 40 political leaders of 27 countries. By scaling the results of a dictionary application, we show how, in comparative terms, liberal or illiberal the leaders present themselves to their national and international audience. In order to gauge whether our liberalness scale reveals meaningful distinctions, we perform a series of validity tests: criterion validity, qualitative hand-coding, unsupervised topic modeling, and network analysis. All tests suggest that our liberalness scale does capture meaningful differences between political regimes despite the large heterogeneity of our data. [ABSTRACT FROM AUTHOR]

Link prediction is one of the most important personalized services in social network platforms. The key point is to predict the probability of the existence of a link between two nodes based on various information in the network. This article combines information of the network structure with the user-generated contents. We propose link prediction indices based on both network structure and topic distribution (NSTD). In contrast to previous literatures, this approach makes full use of the network characteristics, such as homophily, transitivity, clustering, and degree heterogeneity. And we combine these characteristics with topic similarity when constructing indices based on both directly and indirectly connected nodes. Experiment results demonstrate that the proposed method outperforms the previous methods. [ABSTRACT FROM AUTHOR]

Social annotations are valuable resources generated by users on the Web, which encode abundant information on user preferences for certain documents. Social annotation-based information retrieval has been studied in recent years for personalizing search results and fulfilling user information needs. However, since social annotations are complicated and associated with users, documents and tags simultaneously, it remains a great challenge to fully capture the potentially useful information for improving retrieval performance. To meet the challenge, we propose a novel method to integrate social annotations into topic models for personalized document retrieval. Our method first reconstructs candidate documents for a given query using social tags of documents to capture user preferences. The reconstructed documents are tailored to user preferences for achieving better performance. We then generalize the latent Dirichlet allocation-based topic models by considering the relationship among users, social tags and documents from social annotations. The modified topic model optimizes the distribution of latent topics of documents for different users to meet user information needs. Experimental results show that our method can significantly outperform the state-of-the-art baseline models for improving the performance of personalized retrieval. [ABSTRACT FROM AUTHOR]

This research presents a novel aggregating method for constructing an aggregated topic model that is composed of the topics with greater coherence than individual models. When generating a topic model, a number of parameters have to be specified. The resulting topics can be very general or very specific, which depend on the chosen parameters. In this study we investigate the process of aggregating multiple topic models generated using different parameters with a focus on whether combining the general and specific topics is able to increase topic coherence. We employ cosine similarity and Jensen-Shannon divergence to compute the similarity among topics and combine them into an aggregated model when their similarity scores exceed a predefined threshold. The model is evaluated against the standard topics models generated by the latent Dirichlet allocation and Non-negative Matrix Factorisation. Specifically we use the coherence of topics to compare the individual models that create aggregated models against those of the aggregated model and models generated by Non-negative Matrix Factorisation, respectively. The results demonstrate that the aggregated model outperforms those topic models at a statistically significant level in terms of topic coherence over an external corpus. We also make use of the aggregated topic model on social media data to validate the method in a realistic scenario and find that again it outperforms individual topic models. [ABSTRACT FROM AUTHOR]

Literary novels are said to distinguish themselves from other novels through conventions associated with literariness. We investigate the task of predicting the literariness of novels as perceived by readers, based on a large reader survey of contemporary Dutch novels. Previous research showed that ratings of literariness are predictable from texts to a substantial extent using machine learning, suggesting that it may be possible to explain the consensus among readers on which novels are literary as a consensus on the kind of writing style that characterizes literature. Although we have not yet collected human judgments to establish the influence of writing style directly (we use a survey with judgments based on the titles of novels), we can try to analyze the behavior of machine learning models on particular text fragments as a proxy for human judgments. In order to explore aspects of the texts associated with literariness, we divide the texts of the novels in chunks of 2–3 pages and create vector space representations using topic models (Latent Dirichlet Allocation) and neural document embeddings (Distributed Bag-of-Words Paragraph Vectors). We analyze the semantic complexity of the novels using distance measures, supporting the notion that literariness can be partly explained as a deviation from the norm. Furthermore, we build predictive models and identify specific keywords and stylistic markers related to literariness. While genre plays a role, we find that the greater part of factors affecting judgments of literariness are explicable in bag-of-words terms, even in short text fragments and among novels with higher literary ratings. The code and notebook used to produce the results in this paper are available at https://github.com/andreasvc/litvecspace. [ABSTRACT FROM AUTHOR]

The field of evidence synthesis is growing rapidly, with a corresponding increase in the number of software tools and workflows to support the construction of systematic reviews, systematic maps, and meta‐analyses. Despite much progress, however, a number of problems remain, including slow integration of new statistical or methodological approaches into user‐friendly software, low prevalence of open‐source software, and poor integration among distinct software tools. These issues hinder the utility and transparency of new methods to the research community. Here, I present revtools, an R package to support article screening during evidence synthesis projects. It provides tools for the import and deduplication of bibliographic data, screening of articles by title or abstract, and visualization of article content using topic models. The software is entirely open‐source and combines command‐line scripting for experienced programmers with custom‐built user interfaces for casual users, with further methods to support article screening to be added over time. revtools provides free access to novel methods in an open‐source environment and represents a valuable step in expanding the capacity of R to support evidence synthesis projects. [ABSTRACT FROM AUTHOR]