Yelena Mejova

Daniele Liberatore, Kyriaki Kalimeri, Derya Sever et al.

Humanitarian action is accompanied by a mass of reports, summaries, news, and other documents. To guide its activities, important information must be quickly extracted from such free-text resources. Quantities, such as the number of people affected, amount of aid distributed, or the extent of infrastructure damage, are central to emergency response and anticipatory action. In this work, we contribute an annotated dataset for the humanitarian domain for the extraction of such quantitative information, along side its important context, including units it refers to, any modifiers, and the relevant event. Further, we develop a custom Natural Language Processing pipeline to extract the quantities alongside their units, and evaluate it in comparison to baseline and recent literature. The proposed model achieves a consistent improvement in the performance, especially in the documents pertaining to the Dominican Republic and select African countries. We make the dataset and code available to the research community to continue the improvement of NLP tools for the humanitarian domain.

Enrico M. Belliardo, Kyriaki Kalimeri, Yelena Mejova

Geographical location is a crucial element of humanitarian response, outlining vulnerable populations, ongoing events, and available resources. Latest developments in Natural Language Processing may help in extracting vital information from the deluge of reports and documents produced by the humanitarian sector. However, the performance and biases of existing state-of-the-art information extraction tools are unknown. In this work, we develop annotated resources to fine-tune the popular Named Entity Recognition (NER) tools Spacy and roBERTa to perform geotagging of humanitarian texts. We then propose a geocoding method FeatureRank which links the candidate locations to the GeoNames database. We find that not only does the humanitarian-domain data improves the performance of the classifiers (up to F1 = 0.92), but it also alleviates some of the bias of the existing tools, which erroneously favor locations in the Western countries. Thus, we conclude that more resources from non-Western documents are necessary to ensure that off-the-shelf NER systems are suitable for the deployment in the humanitarian sector.

Ivan Decostanzi, Yelena Mejova, Kyriaki Kalimeri

Timely and accurate situational reports are essential for humanitarian decision-making, yet current workflows remain largely manual, resource intensive, and inconsistent. We present a fully automated framework that uses large language models (LLMs) to transform heterogeneous humanitarian documents into structured and evidence-grounded reports. The system integrates semantic text clustering, automatic question generation, retrieval augmented answer extraction with citations, multi-level summarization, and executive summary generation, supported by internal evaluation metrics that emulate expert reasoning. We evaluated the framework across 13 humanitarian events, including natural disasters and conflicts, using more than 1,100 documents from verified sources such as ReliefWeb. The generated questions achieved 84.7 percent relevance, 84.0 percent importance, and 76.4 percent urgency. The extracted answers reached 86.3 percent relevance, with citation precision and recall both exceeding 76 percent. Agreement between human and LLM based evaluations surpassed an F1 score of 0.80. Comparative analysis shows that the proposed framework produces reports that are more structured, interpretable, and actionable than existing baselines. By combining LLM reasoning with transparent citation linking and multi-level evaluation, this study demonstrates that generative AI can autonomously produce accurate, verifiable, and operationally useful humanitarian situation reports.

Corrado Monti, Arthur Capozzi, Yelena Mejova et al.

How do European publics debate a geopolitical crisis on social media, and do they inhabit a shared informational reality? We analyze over 38 million geolocated tweets from 20 European countries during the first eight months of the Russian invasion of Ukraine. Using retweet community detection and stance annotation across six issues, we identify 'hawkish' and 'doveish' opinion clusters present within almost every country studied. We find that structural polarization is driven not by radicalization, but by the exit of casual users. Crucially, whether opposing sides orient to the same events depends on the issue. On pragmatist issues, both sides react to the same high-profile events, forming an agonistic public sphere. Instead, on interpretive issues, they operate as affective publics and counterpublics constructing divergent meanings. We propose conditional publics to describe formations whose relational structure, sharing or fracturing a referential frame, depends on the epistemic character of the debated issue.

Jacopo D'Ignazi, Andreas Kaltenbrunner, Yelena Mejova et al.

Over the last few years, verifying the credibility of information sources has become a fundamental need to combat disinformation. Here, we present a language-agnostic model designed to assess the reliability of web domains as sources in references across multiple language editions of Wikipedia. Utilizing editing activity data, the model evaluates domain reliability within different articles of varying controversiality, such as Climate Change, COVID-19, History, Media, and Biology topics. Crafting features that express domain usage across articles, the model effectively predicts domain reliability, achieving an F1 Macro score of approximately 0.80 for English and other high-resource languages. For mid-resource languages, we achieve 0.65, while the performance of low-resource languages varies. In all cases, the time the domain remains present in the articles (which we dub as permanence) is one of the most predictive features. We highlight the challenge of maintaining consistent model performance across languages of varying resource levels and demonstrate that adapting models from higher-resource languages can improve performance. We believe these findings can assist Wikipedia editors in their ongoing efforts to verify citations and may offer useful insights for other user-generated content communities.

Yelena Mejova, Lydia Manikonda

Unprecedented lockdowns at the start of the COVID-19 pandemic have drastically changed the routines of millions of people, potentially impacting important health-related behaviors. In this study, we use YouTube videos embedded in tweets about diet, exercise and fitness posted before and during COVID-19 to investigate the influence of the pandemic lockdowns on diet and nutrition. In particular, we examine the nutritional profile of the foods mentioned in the transcript, description and title of each video in terms of six macronutrients (protein, energy, fat, sodium, sugar, and saturated fat). These macronutrient values were further linked to demographics to assess if there are specific effects on those potentially having insufficient access to healthy sources of food. Interrupted time series analysis revealed a considerable shift in the aggregated macronutrient scores before and during COVID-19. In particular, whereas areas with lower incomes showed decrease in energy, fat, and saturated fat, those with higher percentage of African Americans showed an elevation in sodium. Word2Vec word similarities and odds ratio analysis suggested a shift from popular diets and lifestyle bloggers before the lockdowns to the interest in a variety of healthy foods, communal sharing of quick and easy recipes, as well as a new emphasis on comfort foods. To the best of our knowledge, this work is novel in terms of linking attention signals in tweets, content of videos, their nutrients profile, and aggregate demographics of the users. The insights made possible by this combination of resources are important for monitoring the secondary health effects of social distancing, and informing social programs designed to alleviate these effects.

Yelena Mejova, Tatiana Gracyk, Ronald E. Robertson

Among the myriad barriers to abortion access, crisis pregnancy centers (CPCs) pose an additional difficulty by targeting women with unexpected or "crisis" pregnancies in order to dissuade them from the procedure. Web search engines may prove to be another barrier, being in a powerful position to direct their users to health information, and above all, health services. In this study we ask, to what degree does Google Search provide quality responses to users searching for an abortion provider, specifically in terms of directing them to abortion clinics (ACs) or CPCs. To answer this question, we considered the scenario of a woman searching for abortion services online, and conducted 10 abortion-related queries from 467 locations across the United States once a week for 14 weeks. Overall, among Google's location results that feature businesses alongside a map, 79.4% were ACs, and 6.9% were CPCs. When an AC was returned, it was the closest known AC location 86.9% of the time. However, when a CPC appeared in a result set, it was the closest one to the search location 75.9% of the time. Examining correlates of AC results, we found that fewer AC results were returned for searches from poorer and rural areas, and those with TRAP laws governing AC facility and clinician requirements. We also observed that Google's performance on our queries significantly improved following a major algorithm update. These results have important implications concerning health access quality and equity, both for individual users and public health policy.

Fabio Poletto, Yunbai Zhang, Andre Panisson et al.

This paper describes in details the design and development of a novel annotation framework and of annotated resources for Internal Displacement, as the outcome of a collaboration with the Internal Displacement Monitoring Centre, aimed at improving the accuracy of their monitoring platform IDETECT. The schema includes multi-faceted description of the events, including cause, quantity of people displaced, location and date. Higher-order facets aimed at improving the information extraction, such as document relevance and type, are proposed. We also report a case study of machine learning application to the document classification tasks. Finally, we discuss the importance of standardized schema in dataset benchmark development and its impact on the development of reliable disaster monitoring infrastructure.

Yelena Mejova, Víctor Suarez-Lledó

National Eating Disorders Association conducts a NEDAwareness week every year, during which it publishes content on social media and news aimed to raise awareness of eating disorders. Measuring the impact of these actions is vital for maximizing the effectiveness of such interventions. This paper is an effort to model the change in behavior of users who engage with NEDAwareness content. We find that, despite popular influencers being involved in the campaign, it is governmental and nonprofit accounts that attract the most retweets. Furthermore, examining the tweeting language of users engaged with this content, we find linguistic categories concerning women, family, and anxiety to be mentioned more within the 15 days after the intervention, and categories concerning affiliation, references to others, and positive emotion mentioned less. We conclude with actionable implications for future campaigns and discussion of the method's limitations.

Yelena Mejova, Kyriaki Kalimeri

Technology has recently been recruited in the war against the ongoing obesity crisis; however, the adoption of Health & Fitness applications for regular exercise is a struggle. In this study, we present a unique demographically representative dataset of 15k US residents that combines technology use logs with surveys on moral views, human values, and emotional contagion. Combining these data, we provide a holistic view of individuals to model their physical exercise behavior. First, we show which values determine the adoption of Health & Fitness mobile applications, finding that users who prioritize the value of purity and de-emphasize values of conformity, hedonism, and security are more likely to use such apps. Further, we achieve a weighted AUROC of .673 in predicting whether individual exercises, and we also show that the application usage data allows for substantially better classification performance (.608) compared to using basic demographics (.513) or internet browsing data (.546). We also find a strong link of exercise to respondent socioeconomic status, as well as the value of happiness. Using these insights, we propose actionable design guidelines for persuasive technologies targeting health behavior modification.

Sina Sajadmanesh, Sina Jafarzadeh, Seyed Ali Osia et al.

Food and nutrition occupy an increasingly prevalent space on the web, and dishes and recipes shared online provide an invaluable mirror into culinary cultures and attitudes around the world. More specifically, ingredients, flavors, and nutrition information become strong signals of the taste preferences of individuals and civilizations. However, there is little understanding of these palate varieties. In this paper, we present a large-scale study of recipes published on the web and their content, aiming to understand cuisines and culinary habits around the world. Using a database of more than 157K recipes from over 200 different cuisines, we analyze ingredients, flavors, and nutritional values which distinguish dishes from different regions, and use this knowledge to assess the predictability of recipes from different cuisines. We then use country health statistics to understand the relation between these factors and health indicators of different nations, such as obesity, diabetes, migration, and health expenditure. Our results confirm the strong effects of geographical and cultural similarities on recipes, health indicators, and culinary preferences across the globe.

Ingmar Weber, Yelena Mejova

To use social media for health-related analysis, one key step is the detection of health-related labels for users. But unlike transient conditions like flu, social media users are less vocal about chronic conditions such as obesity, as users might not tweet "I'm still overweight". As, however, obesity-related conditions such as diabetes, heart disease, osteoarthritis, and even cancer are on the rise, this obese-or-not label could be one of the most useful for studies in public health. In this paper we investigate the feasibility of using profile pictures to infer if a user is overweight or not. We show that this is indeed possible and further show that the fraction of labeled-as-overweight users is higher in U.S. counties with higher obesity rates. Going from public to individual health analysis, we then find differences both in behavior and social networks, for example finding users labeled as overweight to have fewer followers.

Hamed Haddadi, Ferda Ofli, Yelena Mejova et al.

Wearable devices with a wide range of sensors have contributed to the rise of the Quantified Self movement, where individuals log everything ranging from the number of steps they have taken, to their heart rate, to their sleeping patterns. Sensors do not, however, typically sense the social and ambient environment of the users, such as general life style attributes or information about their social network. This means that the users themselves, and the medical practitioners, privy to the wearable sensor data, only have a narrow view of the individual, limited mainly to certain aspects of their physical condition. In this paper we describe a number of use cases for how social media can be used to complement the check-up data and those from sensors to gain a more holistic view on individuals' health, a perspective we call the 360 Quantified Self. Health-related information can be obtained from sources as diverse as food photo sharing, location check-ins, or profile pictures. Additionally, information from a person's ego network can shed light on the social dimension of wellbeing which is widely acknowledged to be of utmost importance, even though they are currently rarely used for medical diagnosis. We articulate a long-term vision describing the desirable list of technical advances and variety of data to achieve an integrated system encompassing Electronic Health Records (EHR), data from wearable devices, alongside information derived from social media data.

Yelena Mejova, Javier Borge-Holthoefer, Ingmar Weber

How are you related to Malawi? Do recent events on the Comoros effect you in any subtle way? Who in your extended social network is in Croatia? We seldom ask ourselves these questions, yet a "long tail" of content beyond our everyday knowledge is waiting to be explored. In this work we propose a recommendation task of creating interest in little-known content by building personalized "bridges" to users. We consider an example task of interesting users in little-known countries, and propose a system which aggregates a user's Twitter profile, network, and tweets to create an interest model, which is then matched to a library of knowledge about the countries. We perform a user study of 69 participants and conduct 11 in-depth interviews in order to evaluate the efficacy of the proposed approach and gather qualitative insight into the effect of multi-faceted use of Twitter on the perception of the bridges. We find the increase in interest concerning little-known content to greatly depend on the pre-existing disposition to it. Additionally, we discover a set of vital properties good bridges must possess, including recency, novelty, emotiveness, and a proper selection of language. Using the proposed approach we aim to harvest the "invisible connections" to make explicit the idea of a "small world" where even a faraway country is more closely connected to you than you might have imagined.

Yelena Mejova, Amy X. Zhang, Nicholas Diakopoulos et al.

How do news sources tackle controversial issues? In this work, we take a data-driven approach to understand how controversy interplays with emotional expression and biased language in the news. We begin by introducing a new dataset of controversial and non-controversial terms collected using crowdsourcing. Then, focusing on 15 major U.S. news outlets, we compare millions of articles discussing controversial and non-controversial issues over a span of 7 months. We find that in general, when it comes to controversial issues, the use of negative affect and biased language is prevalent, while the use of strong emotion is tempered. We also observe many differences across news sources. Using these findings, we show that we can indicate to what extent an issue is controversial, by comparing it with other issues in terms of how they are portrayed across different media.