German Tweets Research Articles

Misleading information in crises: exploring content-specific indicators on Twitter from a user perspective

ABSTRACT Recent crises like the COVID-19 pandemic provoked an increasing appearance of misleading information, emphasising the need for effective user-centered countermeasures as an important field in HCI research. This work investigates how content-specific user-centered indicators can contribute to an informed approach to misleading information. In a threefold study, we conducted an in-depth content analysis of 2382 German tweets on Twitter (now X) to identify topical (e.g. 5G), formal (e.g. links), and rhetorical (e.g. sarcasm) characteristics through manual coding, followed by a qualitative online survey to evaluate which indicators users already use autonomously to assess a tweet's credibility. Subsequently, in a think-aloud study participants qualitatively evaluated the identified indicators in terms of perceived comprehensibility and usefulness. While a number of indicators were found to be particularly comprehensible and useful (e.g. claim for absolute truth and rhetorical questions), our findings reveal limitations of indicator-based interventions, particularly for people with entrenched conspiracy theory views. We derive four implications for digitally supporting users in dealing with misleading information, especially during crises.

A comparison of approaches for imbalanced classification problems in the context of retrieving relevant documents for an analysis.

One of the first steps in many text-based social science studies is to retrieve documents that are relevant for an analysis from large corpora of otherwise irrelevant documents. The conventional approach in social science to address this retrieval task is to apply a set of keywords and to consider those documents to be relevant that contain at least one of the keywords. But the application of incomplete keyword lists has a high risk of drawing biased inferences. More complex and costly methods such as query expansion techniques, topic model-based classification rules, and active as well as passive supervised learning could have the potential to more accurately separate relevant from irrelevant documents and thereby reduce the potential size of bias. Yet, whether applying these more expensive approaches increases retrieval performance compared to keyword lists at all, and if so, by how much, is unclear as a comparison of these approaches is lacking. This study closes this gap by comparing these methods across three retrieval tasks associated with a data set of German tweets (Linder in SSRN, 2017. 10.2139/ssrn.3026393), the Social Bias Inference Corpus (SBIC) (Sap et al. in Social bias frames: reasoning about social and power implications of language. In: Jurafsky et al. (eds) Proceedings of the 58th annual meeting of the association for computational linguistics. Association for Computational Linguistics, p 5477-5490, 2020. 10.18653/v1/2020.aclmain.486), and the Reuters-21578 corpus (Lewis in Reuters-21578 (Distribution 1.0). [Data set], 1997. http://www.daviddlewis.com/resources/testcollections/reuters21578/). Results show that query expansion techniques and topic model-based classification rules in most studied settings tend to decrease rather than increase retrieval performance. Active supervised learning, however, if applied on a not too small set of labeled training instances (e.g.1000 documents), reaches a substantially higher retrieval performance than keyword lists.

The interaction between human demand and urban greenspace supply for promoting positive emotions with sentiment analysis from twitter

Urban greenspace (UGS) plays an essential role in providing benefits to human well-being in cities. Understanding how to promote positive emotions is vital for planning and designing a UGS supply that satisfies human demand. However, little is known about the impact on human emotions of the interaction of cross-cultural demands and different greenspace supplies. This study explored different human emotions about UGS from a cross-linguistics perspective using social media data (SMD). Sentiment analysis was conducted with geolocated Twitter data from 26 UGSs in Berlin to acquire sentiment value and tweet number of tweets. The sentiments of English and German tweets in four types of UGSs were compared, and the correlations with 11 physical and activity landscape characteristics were identified. The results demonstrate that (1) sentiment value and tweet number of tweets were distinctive in the 26 greenspaces, showing different emotions responding to the different types of UGS; (2) the cross-linguistic demands were different in the comparison of English and German tweets, with the highest sentiment values in gardens and parks, respectively; and (3) the sentiment of the all, English and German tweets was respectively correlated with open space, interesting plants and swimming infrastructure. The activity landscape made the highest contributions to positive emotions even with cross-cultural differences. The results of the study suggest that human emotions can indicate whether the UGS supply meets the human demand and that specific landscape characteristics can enhance positive emotions to maintain human demand in cross-cultural background, especially considering the increasing attention to immigrants and natives. Thus, human emotions identify the interaction between UGS supply and human demand based on SMD to improve UGS outcomes for urban sustainability and public well-being.

Collective response to the health crisis among German Twitter users: A structural topic modeling approach

We used structural topic modeling to analyze over 800,000 German tweets about COVID-19 to answer the questions: What patterns emerge in tweets as a response to a health crisis? And how do topics discussed change over time? The study leans on the goals associated with the health information seeking (GAINS) model, discerning whether a post aims at tackling and eliminating the problem (i.e., problem-focused) or managing the emotions (i.e., emotion-focused); whether it strives to maximize positive outcomes (promotion focus) or to minimize negative outcomes (prevention focus). The findings indicate four clusters salient in public reactions: 1) “Understanding” (problem-promotion); 2) “Action planning” (problem-prevention); 3) “Hope” (emotion-promotion) and 4) “Reassurance” (emotion-prevention). Public communication is volatile over time, and a shift is evidenced from self-centered to community-centered topics within 4.5 weeks. Our study illustrates social media text mining's potential to quickly and efficiently extract public opinions and reactions. Monitoring fears and trending topics enable policymakers to rapidly respond to deviant behavior, like resistive attitudes toward containment measures or deteriorating physical health. Healthcare workers can use the insights to provide mental health services for battling anxiety or extensive loneliness from staying home.

Efficient and Reliable Geocoding of German Twitter Data to Enable Spatial Data Linkage to Official Statistics and Other Data Sources

More and more, social scientists are using (big) digital behavioral data for their research. In this context, the social network and microblogging platform Twitter is one of the most widely used data sources. In particular, geospatial analyses of Twitter data are proving to be fruitful for examining regional differences in user behavior and attitudes. However, ready-to-use spatial information in the form of GPS coordinates is only available for a tiny fraction of Twitter data, limiting research potential and making it difficult to link with data from other sources (e.g., official statistics and survey data) for regional analyses. We address this problem by using the free text locations provided by Twitter users in their profiles to determine the corresponding real-world locations. Since users can enter any text as a profile location, automated identification of geographic locations based on this information is highly complicated. With our method, we are able to assign over a quarter of the more than 866 million German tweets collected to real locations in Germany. This represents a vast improvement over the 0.18% of tweets in our corpus to which Twitter assigns geographic coordinates. Based on the geocoding results, we are not only able to determine a corresponding place for users with valid profile locations, but also the administrative level to which the place belongs. Enriching Twitter data with this information ensures that they can be directly linked to external data sources at different levels of aggregation. We show possible use cases for the fine-grained spatial data generated by our method and how it can be used to answer previously inaccessible research questions in the social sciences. We also provide a companion R package, nutscoder, to facilitate reuse of the geocoding method in this paper.

From Hostile to Benevolent Ageism: Polarizing Attitudes Toward Older Adults in German COVID-19-Related Tweets

Background and ObjectivesPrevious studies have linked COVID-19 to a rise in ageism. While a growing body of research examined hostile ageism during the pandemic, benevolent ageism received less attention. Drawing on the stereotype content theory and the classic tripartite model of attitudes, the current study explored how benevolent and hostile ageism are reflected in the cognitive, affective, and behavioural dimensions of attitudes towards older adults in German COVID-19 related tweets. The study examined the most prevalent attitudes as well as changes in prevalence between the first and second lockdown period in Germany.Research Design and Methods792 German tweets concerning COVID-19 and ageing were collected and coded using Mayring’s qualitative content analysis with a dominantly inductive approach. Quantitative methods were used to identify the most prevalent subthemes as well as changes in prevalence.ResultsThe coding resulted in 21 subthemes. Most tweets (60.73%) contained either hostile or benevolent ageist attitudes, with benevolent ageism being more prevalent. The top 5 subthemes in terms of prevalence and reach contained several opposing attitudes, such as devaluation and opposing devaluation. The chi-square tests revealed a shift from a promotion to an evaluation of COVID-19 related policies between the two lockdowns.Discussion and ImplicationsResults highlight social media’s polarising effect and its potential contribution to both hostile and benevolent ageism in the context of COVID-19 in Germany. Results indicate the need to consider the adverse effects of benevolent ageism and use of chronological age as risk factor, when designing COVID-19 related policies.

Central bank communication with non-experts – A road to nowhere?

Central banks have intensified their communication with non-experts – an endeavour which some argue will fail. This paper studies English and German tweets about the ECB to show that its communication is received by non-experts, i.e. is not a road to nowhere. Following ECB communications, tweets often primarily relay information, become more factual and the views expressed more moderate and homogeneous. Some communications, such as Mario Draghi's “Whatever it takes”, trigger a divergence in views. Also, tweets with negative, stronger or more subjective views are more likely retweeted, liked or replied to. Thus, Twitter also constitutes a platform for controversial discussions.

Supporting customer-oriented marketing with artificial intelligence: automatically quantifying customer needs from social media

The elicitation and monitoring of customer needs is an important task for businesses, allowing them to design customer-centric products and services and control marketing activities. While there are different approaches available, most lack in automation, scalability and monitoring capabilities. In this work, we demonstrate the feasibility towards an automated prioritization and quantification of customer needs from social media data. To do so, we apply a supervised machine learning approach on the example of previously labeled Twitter data from the domain of e-mobility. We descriptively code over 1000 German tweets and build eight distinct classification models, so that a resulting artifact can independently determine the probabilities of a tweet containing each of the eight previously defined needs. To increase the scope of application, we deploy the machine learning models as part of a web service for public use. The resulting artifact can provide valuable insights for need elicitation and monitoring when analyzing user-generated content on a large scale.

The Automatic Identification of German Afd Supporters on Twitter: The Vanellus-Follower-Classification-Index Vafci.

Vanellus (the peewit) is a twitter-bot developed by Christian “Chrilly” Donninger. The bot downloads all German tweets in real-time and stores them in a special-purpose and efficient database. The database can be searched and analyzed according a host of criterion. This article describes the automated identification of the German AfD – a new right-wing-populist-party – supporters with the Vanellus Follower Classification Index VaFCI. The VaFCI is not restricted to this sort of classification. The supporters of the AfD were selected, because the AfD is rather present and active on Twitter and there is a clear polarization between supporters and opponents. For a human it is relatively easy to classify a user in Pro or Contra. The VaFCI misses a few – for a human – obvious AfD supporters. But a classified supporter is almost always indeed one.

Measuring the Influence of German Members of Parliament on Twitter: The Vanellus-Twitter-Index VaTI

Vanellus (the peewit) is a twitter-bot developed by Christian “Chrilly” Donninger. The bot downloads in real-time all German tweets and stores them in a special-purpose and efficient database. The database can be searched and analyzed according a host of criterion. This article describes the Vanellus-Twitter-Index VaTI for the German MdBs (Mitglieder des Bundestages/Members of Parliament) and for the leaders of the Austrian parties. The VaTI measures the real Twitter-footprint of users. It is much more informative than usual the measure “number of followers”.

Information Propagation Speed and Patterns in Social Networks: A Case Study Analysis of German Tweets

Information Propagation Speed and Patterns in Social Networks: A Case Study Analysis of German Tweets

Microblogging during the European Floods 2013

Social media is becoming more and more important in crisis management. However its analysis by emergency services still bears unaddressed challenges and the majority of studies focus on the use of social media in the USA. In this paper German tweets of the European Flood 2013 are therefore captured and analyzed using descriptive statistics, qualitative data coding, and computational algorithms. The authors' work illustrates that this event provided sufficient German traffic and geo-locations as well as enough original data (not derivative). However, up-to-date Named Entity Recognizer (NER) with German classifier could not recognize German rivers and highways satisfactorily. Furthermore the authors' analysis revealed pragmatic (linguistic) barriers resulting from irony, wordplay, and ambiguity, as well as in retweet-behavior. To ease the analysis of data they suggest a retweet ratio, which is illustrated to be higher with important tweets and may help selecting tweets for mining. The authors argue that existing software has to be adapted and improved for German language characteristics, also to detect markedness, seriousness and truth.