Long-form Text Research Articles

HyperMatch: long-form text matching via hypergraph convolutional networks

HyperMatch: long-form text matching via hypergraph convolutional networks

Computing Short Films Using Language-Guided Diffusion and Vocoding Through Virtual Timelines of Summaries

Language-guided generative models are increasingly used in audiovisual production. Image diffusion allows for the development of video sequences and some of its coordination can be established by text prompts. This research automates a video production pipeline leveraging CLIP-guidance with longform text inputs and a separate text-to-speech system. We introduce a method for producing frame-accurate video and audio summaries using a virtual timeline and document a set of video outputs with diverging parameters. Our approach was applied in the production of the film Irreplaceable Biography and contributes to a future where multimodal generative architectures are set as underlying mechanisms to establish visual sequences in time. We contribute to a practice where language modelling is part of a shared and learned representation which can support professional video production, specifically used as a vehicle throughout the composition process as potential videography in physical space.

Applications of Natural Language Processing to Geoscience Text Data and Prospectivity Modeling

Geological maps are powerful models for visualizing the complex distribution of rock types through space and time. However, the descriptive information that forms the basis for a preferred map interpretation is typically stored in geological map databases as unstructured text data that are difficult to use in practice. Herein we apply natural language processing (NLP) to geoscientific text data from Canada, the U.S., and Australia to address that knowledge gap. First, rock descriptions, geological ages, lithostratigraphic and lithodemic information, and other long-form text data are translated to numerical vectors, i.e., a word embedding, using a geoscience language model. Network analysis of word associations, nearest neighbors, and principal component analysis are then used to extract meaningful semantic relationships between rock types. We further demonstrate using simple Naive Bayes classifiers and the area under receiver operating characteristics plots (AUC) how word vectors can be used to: (1) predict the locations of “pegmatitic” (AUC = 0.962) and “alkalic” (AUC = 0.938) rocks; (2) predict mineral potential for Mississippi-Valley-type (AUC = 0.868) and clastic-dominated (AUC = 0.809) Zn-Pb deposits; and (3) search geoscientific text data for analogues of the giant Mount Isa clastic-dominated Zn-Pb deposit using the cosine similarities between word vectors. This form of semantic search is a promising NLP approach for assessing mineral potential with limited training data. Overall, the results highlight how geoscience language models and NLP can be used to extract new knowledge from unstructured text data and reduce the mineral exploration search space for critical raw materials.

Increasing Women's Knowledge about HPV Using BERT Text Summarization: An Online Randomized Study.

Despite the availability of online educational resources about human papillomavirus (HPV), many women around the world may be prevented from obtaining the necessary knowledge about HPV. One way to mitigate the lack of HPV knowledge is the use of auto-generated text summarization tools. This study compares the level of HPV knowledge between women who read an auto-generated summary of HPV made using the BERT deep learning model and women who read a long-form text of HPV. We randomly assigned 386 women to two conditions: half read an auto-generated summary text about HPV (n = 193) and half read an original text about HPV (n = 193). We administrated measures of HPV knowledge that consisted of 29 questions. As a result, women who read the original text were more likely to correctly answer two questions on the general HPV knowledge subscale than women who read the summarized text. For the HPV testing knowledge subscale, there was a statistically significant difference in favor of women who read the original text for only one question. The final subscale, HPV vaccination knowledge questions, did not significantly differ across groups. Using BERT for text summarization has shown promising effectiveness in increasing women’s knowledge and awareness about HPV while saving their time.

Data-to-text Generation with Variational Sequential Planning

Abstract We consider the task of data-to-text generation, which aims to create textual output from non-linguistic input. We focus on generating long-form text, that is, documents with multiple paragraphs, and propose a neural model enhanced with a planning component responsible for organizing high-level information in a coherent and meaningful way. We infer latent plans sequentially with a structured variational model, while interleaving the steps of planning and generation. Text is generated by conditioning on previous variational decisions and previously generated text. Experiments on two data-to-text benchmarks (RotoWire and MLB) show that our model outperforms strong baselines and is sample-efficient in the face of limited training data (e.g., a few hundred instances).

Reading people

Empathy allows us to understand suffering better, and fostering it, especially in young people, is crucial for societies to thrive. The reduction in the amount literature and long-form text that young people read could potentially have implications for their levels of empathy.

New Approaches to Islamic Law and the Documentary Record before 1500

Marina Rustow notes how prevalent scholarly attention is to long-form texts of Islamic law—attention that she argues, comes at the expense of studying Islamic legal documents in a sufficient manner. Study of the documents is an indispensable enterprise if we are to fully understand “how law worked in practice.” In view of what we know to have been “heaps” of documents produced by Muslim judges and notaries, Rustow underscores how particularly noticeable a disjuncture there is between those documents and the long-form texts. Moreover, scholars often skip over and thus fail to avail themselves of the utility of documents in adding texture to social and legal history. She cautions social historians against “pseudo-knowledge,” that is, the temptation to overlook complex factors, usually embedded in legal documents, that render our otherwise tame scholarly perception of the past truer but more “unruly.” In the end, her invitation to join her in the study of documents and thereby improve the state of Islamic legal history is terse and timely: “Please go find yourself some documents.”

Pinnacles in Long-form Literary Texts: Cross-textual Evidence for the Pervasiveness of Megametaphorical Expression

Various scholars have recognized how literary authors use mega-metaphor (scattered metaphorical domain references) to unobtrusively shape their narratives. Corpus analysis proves useful not only in identifying the presence of such metaphorical systems in long-form texts but also in assessing their relative prominence across multiple texts. This article will examine 50 randomly selected literary narratives so as to discern the extent to which the relatively uncommon but metaphorically replete word pinnacle contributes surreptitious metaphorical meaning to the literary texts in which it is found. Results of the study confirm that the word “pinnacle,” when it is used at all, is nearly always placed so as to highlight climactic scenes or emphasize key turning points in protagonist character development. The fact that metaphorical lines of interpretation relating to certain words and phrases may be detected, not only by way of critical intuition but also through the electronic searching of multiple text corpora, demonstrates the value of cross-textual analysis strategies in certain cases. It also hints that megametaphor, rather than being a rare and idiosyncratic type of literary artifice, may be more prevalent than has been previously acknowledged.

A Toxic Blend: Assessing the Effects of Cross-Source Media Coverage of Flavored E-Cigarettes on Youth and Young Adult Perceptions

Flavored e-cigarettes have received high attention across social and news media. How does exposure to e-cigarette flavors across multiple sources in the media environment influence youth e-cigarette perceptions? To address this question, we identified e-cigarette flavor mentions on 24.3 million Twitter posts and 11,691 longform texts (newspapers, broadcast news, and websites) disseminated over 3 years (2014–2017). During the same period, we measured e-cigarette beliefs through a nationally representative randomly sampled rolling survey of 13–26-year-olds (N = 4,470, 1013 days). We estimated the association between flavor-specific content on Twitter and longform sources in the 28 days prior to each survey date and perceptions that e-cigarettes taste good. The interaction of coverage on Twitter and longform sources was significantly associated with more favorable perceptions of e-cigarette taste (OR = 1.21; 95% CI, 1.04–1.41); the main effects of each source type were not significant. This study presents a novel approach to evaluating the effects of cross-source coverage in today’s complex media landscape and may strengthen claims for media influence on e-cigarette use.

Toward an Aggregate, Implicit, and Dynamic Model of Norm Formation: Capturing Large-Scale Media Representations of Dynamic Descriptive Norms Through Automated and Crowdsourced Content Analysis.

Media content can shape people's descriptive norm perceptions by presenting either population-level prevalence information or descriptions of individuals' behaviors. Supervised machine learning and crowdsourcing can be combined to answer new, theoretical questions about the ways in which normative perceptions form and evolve through repeated, incidental exposure to normative mentions emanating from the media environment. Applying these methods, this study describes tobacco and e-cigarette norm prevalence and trends over 37months through an examination of a census of 135,764 long-form media texts, 12,262 popular YouTube videos, and 75,322,911 tweets. Long-form texts mentioned tobacco population norms (4-5%) proportionately less often than e-cigarette population norms (20%). Individual use norms were common across sources, particularly YouTube (tobacco long-form: 34%; Twitter: 33%; YouTube: 88%; e-cigarette long form: 17%; Twitter: 16%; YouTube: 96%). The capacity to capture aggregated prevalence and temporal dynamics of normative media content permits asking population-level media effects questions that would otherwise be infeasible to address.

Combining Crowd-Sourcing and Automated Content Methods to Improve Estimates of Overall Media Coverage: Theme Mentions in E-cigarette and Other Tobacco Coverage

Exposure to media content can shape public opinions about tobacco. Accurately describing content is a first step to showing such effects. Historically, content analyses have hand-coded tobacco-focused texts from a few media sources which ignored passing mention coverage and social media sources, and could not reliably capture over-time variation. By using a combination of crowd-sourced and automated coding, we labeled the population of all e-cigarette and other tobacco-related (including cigarettes, hookah, cigars, etc.) ‘long-form texts’ (focused and passing coverage, in mass media and website articles) and social media items (tweets and YouTube videos) collected May 2014-June 2017 for four tobacco control themes. Automated coding of theme coverage met thresholds for item-level precision and recall, event validation, and weekly-level reliability for most sources, except YouTube. Health, Policy, Addiction and Youth themes were frequent in e-cigarette long-form focused coverage (44%-68%), but not in long-form passing coverage (5%-22%). These themes were less frequent in other tobacco coverage (long-form focused (13-32%) and passing coverage (4-11%)). Themes were infrequent in both e-cigarette (1-3%) and other tobacco tweets (2-4%). Findings demonstrate that passing e-cigarette and other tobacco long-form coverage and social media sources paint different pictures of theme coverage than focused long-form coverage. Automated coding also allowed us to code the amount of data required to estimate reliable weekly theme coverage over three years. E-cigarette theme coverage showed much more week-to-week variation than did other tobacco coverage. Automated coding allows accurate descriptions of theme coverage in passing mentions, social media, and trends in weekly theme coverage.

What people study when they study Tumblr

PurposeSince its launch in 2007, research has been carried out on the popular social networking website Tumblr. The purpose of this paper is to identify published Tumblr-based research, classify it to understand approaches and methods, and provide methodological recommendations for others.Design/methodology/approachResearch regarding Tumblr was identified. Following a review of the literature, a classification scheme was adapted and applied, to understand research focus. Papers were quantitatively classified using open coded content analysis of method, subject, approach, and topic.FindingsThe majority of published work relating to Tumblr concentrates on conceptual issues, followed by aspects of the messages sent. This has evolved over time. Perceived benefits are the platform’s long-form text posts, ability to track tags, and the multimodal nature of the platform. Severe research limitations are caused by the lack of demographic, geo-spatial, and temporal metadata attached to individual posts, the limited Advanced Programming Interface, restricted access to data, and the large amounts of ephemeral posts on the site.Research limitations/implicationsThis study focusses on Tumblr: the applicability of the approach to other media is not considered. The authors focus on published research and conference papers: there will be book content which was not found using the method. Tumblr as a platform has falling user numbers which may be of concern to researchers.Practical implicationsThe authors identify practical barriers to research on the Tumblr platform including lack of metadata and access to big data, explaining why Tumblr is not as popular as Twitter in academic studies.Social implicationsThis paper highlights the breadth of topics covered by social media researchers, which allows us to understand popular online platforms.Originality/valueThere has not yet been an overarching study to look at the methods and purpose of those who study Tumblr. The authors identify Tumblr-related research papers from the first appearing in 2011 July until 2015 July. The classification derived here provides a framework that can be used to analyse social media research, and in which to position Tumblr-related work, with recommendations on benefits and limitations of the platform for researchers.

Weiss/Sartre: Cold War Stellungnahme and the Poetic–Political Either/Or in Peter Weiss's Final Swedish Novel

This essay studies the emergence of the concept of Stellungnahme in the prose of Peter Weiss. The last long-form text Weiss wrote in Swedish, the unpublished 1956 novel Die Situation, represents a serious consideration of the literary theory of Jean-Paul Sartre, whose arguments about the different efficacies of prose and poetry inform much of Weiss's early thinking about the relationship between aesthetics and politics and about his own place in the world of letters and ideas. Connecting Weiss's early encounters with Sartre's critical models to his frequent return to the originary scene of Stellungnahme in his German-language autobiographical writing, this essay recovers Sartrean existentialism's early importance for Peter Weiss's thought, finally reflecting on the persistence of an existentialist vocabulary of situation in Weiss's later, better-known work.

Digital reading spaces: How expert readers handle books, the Web and electronic paper

This article focuses on changing reading characteristics and presents a study among a group of expert readers. Considering technological bases of reading and applying corporeal and material perspectives, the study examines manners in which proficient readers handle printed and digital texts, attempting to explain differences in digital and paper-based reading. Based on findings, the article reflects on how long-form text can be productively transferred into the digital reading space.