Fine-grained Analysis Research Articles

Estimating the legibility of international borders.

Researchers in the social sciences are interested in the consequences of institutions, increasingly on a global scale. Institutions that may be negotiated between states can have consequences at a microlevel, as local populations adjust their expectations and ultimately even their behavior to take institutional rules into account. However, large-scale fine-grained analyses that test for the complex evidence of such institutions locally are rare. This article focuses on a key institution: International borders. Using computer vision techniques, we show that it is possible to produce a geographically specific, validated, and replicable way to characterize border legibility, by which we mean the ability to visually detect the presence of an international border in physical space. We develop and compare computer vision techniques to automatically estimate legibility scores for 627,656 imagery tiles from virtually every border in the world. We evaluate statistical and data-driven computer vision methods, finding that fine-tuning pretrained visual recognition models on a small set of human judgments allows us to produce local legibility scores globally that align well with human notions of legibility. Finally, we interpret these scores as useful approximations of states' border orientations, a concept that prior literature has used to capture the visible investments states make in border areas to maintain jurisdictional authority territorially. We validate our measurement strategy using both human judgments and five nomological validation indicators.

Neural and behavioral similarity-driven tuning curves for manipulable objects

Abstract In our daily activities, we encounter numerous objects that we successfully distinguish and recognize within a fraction of a second. This holds for coarse distinctions (e.g., cat vs. hammer) but also for more challenging distinctions that require fine-grain analysis (e.g., cat vs. dog). The efficiency of this recognition depends on how the brain organizes object-related information. While several attempts have focused on unraveling large-scale organization principles, research on fine-grained knowledge organization is rather limited. Here, we explored the fine-grain organization of object knowledge and investigated whether manipulable objects are organized and represented in terms of their similarity. To accomplish this, different groups of individuals participated in a behavioral and fMRI release from adaptation experiment. Adaptation was induced by presenting different exemplars of a particular object, and release from adaptation was elicited by the presentation of a deviant object. The relationship between adaptation and deviant objects was manipulated into four levels of similarity, measured by feature overlap between these objects. Our findings revealed that increasing object similarity provoked slower reaction times and weaker fMRI release from adaptation. Specifically, we identified similarity-driven tuning curves for the release from adaptation in the medial fusiform, collateral sulcus, parahippocampal gyri, lingual gyri, lateral occipital complex, and occipito-parietal cortex. These results suggest that the processing and representation of objects in the brain and our ability to perform fine discriminations between objects reflect real-world object similarity in a relatively parametric manner.

A novel large language model enhanced joint learning framework for fine-grained sentiment analysis on drug reviews

A novel large language model enhanced joint learning framework for fine-grained sentiment analysis on drug reviews

From speech signal to syntactic structure: A computational implementation

This paper presents a new computational implementation bridging several modules of grammar from phonetics to phonology to syntax. The system takes as input a speech signal annotated with syllables, interprets the phonetic data in phonological/prosodic terms, matches the data against a lexicon and makes the results available to a linguistically deep computational grammar. The system is showcased by means of syntactically ambiguous structures in German which can be disambiguated based on prosodic constituency information. A system evaluation with the German data showed good results for this new combination of automatic speech signal analysis and computational grammars, which takes a significant step towards a linguistically fine-grained computational analysis and hence towards real automatic speech understanding.

Heat and Hate: Climate Security and Farmer-Herder Conflicts in Africa

Abstract We investigate the impact of climate shocks on farmer-herder violence, using geolocalized data on conflict events for Africa, 1997-2014. We find that a one degree temperature increase yields a 54% increase in conflict probability in mixed areas populated by both farmers and herders (versus 17% in non-mixed areas). Quantifying the impact of projected climate change in 2040, we find that -when factoring in the magnifying effect of mixed settlements- annual conflicts are predicted to rise by a third. Drawing on a fine-grained analysis of groups' mobility patterns, we show that resource competition is a major driver of farmer-herder violence.

Aspect category sentiment analysis based on pre-trained BiLSTM and syntax-aware graph attention network

Aspect Category Sentiment Analysis (ACSA) is a fine-grained sentiment analysis task aimed at predicting the sentiment polarity associated with aspect categories within a sentence.Most existing ACSA methods are based on a given aspect category to locate sentiment words related to it. When irrelevant sentiment words have semantic meaning for the given aspect category, it may cause the problem that sentiment words cannot be matched with aspect categories. To address the aforementioned issue, this paper proposes a novel approach for ACSA utilizing pre-trained Bidirectional Long Short-Term Memory (BiLSTM) and syntax-aware graph attention network. To address the issue of insufficient existing annotated datasets, a method of using transfer learning is proposed. Firstly, the BiLSTM model is used to pre-train on the document-level sentiment analysis dataset, and the obtained pre-training parameters are transferred to the aspect-level task model. Then, a syntax-aware graph attention network model is proposed to make full use of the syntactic structure and semantic information in the text, and combine the knowledge learned in pre-training to achieve the ACSA task. The performance evaluation of this method is carried out on five user comment text datasets, and the comprehensive ablation experiments prove that this method performs best compared with baseline models.

Surgical video workflow analysis via visual-language learning

Surgical video workflow analysis has made intensive development in computer-assisted surgery by combining deep learning models, aiming to enhance surgical scene analysis and decision-making. However, previous research has primarily focused on coarse-grained analysis of surgical videos, e.g., phase recognition, instrument recognition, and triplet recognition that only considers relationships within surgical triplets. In order to provide a more comprehensive fine-grained analysis of surgical videos, this work focuses on accurately identifying triplets <instrument, verb, target> from surgical videos. Specifically, we propose a vision-language deep learning framework that incorporates intra- and inter- triplet modeling, termed I2TM, to explore the relationships among triplets and leverage the model understanding of the entire surgical process, thereby enhancing the accuracy and robustness of recognition. Besides, we also develop a new surgical triplet semantic enhancer (TSE) to establish semantic relationships, both intra- and inter-triplets, across visual and textual modalities. Extensive experimental results on surgical video benchmark datasets demonstrate that our approach can capture finer semantics, achieve effective surgical video understanding and analysis, with potential for widespread medical applications.

The perceptual primacy of feeling: Affectless visual machines explain a majority of variance in human visually evoked affect

Looking at the world often involves not just seeing things, but feeling things. Modern feedforward machine vision systems that learn to perceive the world in the absence of active physiology, deliberative thought, or any form of feedback that resembles human affective experience offer tools to demystify the relationship between seeing and feeling, and to assess how much of visually evoked affective experiences may be a straightforward function of representation learning over natural image statistics. In this work, we deploy a diverse sample of 180 state-of-the-art deep neural network models trained only on canonical computer vision tasks to predict human ratings of arousal, valence, and beauty for images from multiple categories (objects, faces, landscapes, art) across two datasets. Importantly, we use the features of these models without additional learning, linearly decoding human affective responses from network activity in much the same way neuroscientists decode information from neural recordings. Aggregate analysis across our survey, demonstrates that predictions from purely perceptual models explain a majority of the explainable variance in average ratings of arousal, valence, and beauty alike. Finer-grained analysis within our survey (e.g. comparisons between shallower and deeper layers, or between randomly initialized, category-supervised, and self-supervised models) point to rich, preconceptual abstraction (learned from diversity of visual experience) as a key driver of these predictions. Taken together, these results provide further computational evidence for an information-processing account of visually evoked affect linked directly to efficient representation learning over natural image statistics, and hint at a computational locus of affective and aesthetic valuation immediately proximate to perception.

Ancient DNA reveals reproductive barrier despite shared Avar-period culture.

After a long-distance migration, Avars with Eastern Asian ancestry arrived in Eastern Central Europe in 567 to 568 CE and encountered groups with very different European ancestry1,2. We used ancient genome-wide data of 722 individuals and fine-grained interdisciplinary analysis of large seventh- to eighth-century CE neighbouring cemeteries south of Vienna (Austria) to address the centuries-long impact of this encounter1,2. We found that even 200 years after immigration, the ancestry at one site (Leobersdorf) remained dominantly East Asian-like, whereas the other site (Mödling) shows local, European-like ancestry. These two nearby sites show little biological relatedness, despite sharing a distinctive late-Avar culture3,4. We reconstructed six-generation pedigrees at both sites including up to 450 closely related individuals, allowing per-generation demographic profiling of the communities. Despite different ancestry, these pedigrees together with large networks of distant relatedness show absence of consanguinity, patrilineal pattern with female exogamy, multiple reproductive partnerships (for example, levirate) and direct correlation of biological connectivity with archaeological markers of social status. The generation-long genetic barrier was maintained by systematically choosing partners with similar ancestry from other sites in the Avar realm. Leobersdorf had more biological connections with the Avar heartlands than with Mödling, which is instead linked to another site from the Vienna Basin with European-like ancestry. Mobility between sites was mostly due to female exogamy pointing to different marriage networks as the main driver of the maintenance of the genetic barrier.

Understanding emergent leader–follower patterns and the role of gender in teams: a micro-temporal account

Purpose By investigating the dynamics of leader and follower behavior during small group interactions, we provide insights into the behavioral patterns that give rise to leadership emergence. We also identify gender-related differences in these behavior patterns that may explain the persistent gap in emergent leadership ascriptions between men and women.Design/methodology/approach We video-recorded verbal interactions of 34 zero-history three-person teams collaborating on a task in the laboratory. One team member was a confederate (male vs. female) trained to show emergent leader behavior. To quantify verbal interaction patterns and examine to what extent these team dynamics depend on the confederate’s gender, we conducted a fine-grained interaction analysis of utterances over the interaction period.FindingsWe show that leadership claims by one team member evoked subsequent granting behavior in another team member. The more individuals’ claims were granted (counter-claimed) by others, the higher (lower) their level of ascribed emergent leadership. Claims uttered by male or female confederates were equally likely to be granted by team members. However, female confederate leadership claims elicited more counterclaims.Originality/value Findings highlight the importance of considering leader–follower interaction patterns for the discussion around gender differences in leadership processes.

Intergenerational Relationships and Family Support: Implications for Health and Wellbeing.

This review summarizes research projects supported by the National Institute on Aging (NIA) that have contributed scholarship on intergenerational relationships and support provided to older adults that frequently precedes, and is often complementary to, intensive caregiving. We review NIA-supported projects that have almost exclusively focused on intergenerational relationships and involved primary data collections, and others making use of omnibus aging and family studies that have allowed a variety of investigations on this topic. Where the former set of studies has generated deeply phenotypic analyses-comprehensive fine-grained analyses of relational data in specialized samples-the latter set has focused on analyses of secondary data, often from national samples which include information on intergenerational relationships. Early research funded by NIA addressed the factors underlying cohesion across generations, including the Longitudinal Study of Generations. Subsequent studies shed light on the dynamics of ties between siblings or across multiple generations in navigating support and affection. Studies have revealed important information about transfers of tangible and non-tangible resources, loss of parent or child, and diversity by race, ethnicity, and gender. NIA funding has made important inroads in understanding a relationship that is of primary importance in individuals' lives for their health and well-being.

Emotional pictures in the brain and their interaction with the task: A fine-grained fMRI coordinate-based meta-analysis study.

The impacting research on emotions of the last decades was carried out with different methods. The most popular was based on the use of a validated sample of slides, the International Affective Pictures System (IAPS), divided mainly into pleasant, neutral and unpleasant categories, and on fMRI as a measure of brain activation induced by these stimuli. With the present coordinate-based meta-analysis (CBMA) based on ALE approach, we aimed to unmask the main brain networks involved in the contrast of pleasant vs. neutral and unpleasant vs. neutral IAPS slide categories. Furthermore, we included studies employing both IAPS and non-IAPS (but analogously validated) pictures, a condition termed as IAPS EXTENDED. After selecting 97 papers published in the 2000-2023 interval, the planned contrasts were analyzed by also considering their interaction with the Load factor of the concomitant task, which comprised the conditions: No Load (passive viewing), Low-Load tasks and High-Load tasks. We analyzed a total of 152 experiments (106 focusing on the negative vs. neutral contrast; 46 reporting positive vs. neutral contrasts). We additionally performed conjunction and contrast analyses. Results confirmed outcomes of past meta-analyses on the involvement of a number of cortical and subcortical paralimbic and limbic regions during unpleasant picture processing, but the increase of the pubblications on this topic in last years, together with a more fine-grained analysis, allowed us to find also the involvement of additional areas, such as the right middle frontal gyrus, left inferior frontal gyrus (BA 9), posterior cingulate, and left inferior parietal lobule. Concerning passive viewing and low-load tasks, a clear frontal asymmetry emerged with greater right prefrontal activation (BA9) to unpleasant vs. pleasant pictures, whereas, during No Load tasks only, left frontal dominance to pleasant vs. unpleasant stimuli was found (BA13). The unpleasant vs. neutral comparison on High-Load tasks (the pleasant condition had an insufficient sample size) revealed a specific lateralization of several areas of the right hemisphere (STG-BA 38, MFG-BA 46, FG-BA 37), whereas, in the other load conditions, the inferior frontal gyrus was right lateralized, but the main activated regions were bilateral or left lateralized. Results are discussed considering the effects of both valence and task/load variables, and the involvement of hippocampus/amygdala, hemispheric asymmetries of emotions, the occipito-temporal areas, several sub-regions of the prefrontal/orbitofrontal cortex, and an extended motor network.

Smartboard: Visual Exploration of Team Tactics with LLM Agent.

Tactics play an important role in team sports by guiding how players interact on the field. Both sports fans and experts have a demand for analyzing sports tactics. Existing approaches allow users to visually perceive the multivariate tactical effects. However, these approaches require users to experience a complex reasoning process to connect the multiple interactions within each tactic to the final tactical effect. In this work, we collaborate with basketball experts and propose a progressive approach to help users gain a deeper understanding of how each tactic works and customize tactics on demand. Users can progressively sketch on a tactic board, and a coach agent will simulate the possible actions in each step and present the simulation to users with facet visualizations. We develop an extensible framework that integrates large language models (LLMs) and visualizations to help users communicate with the coach agent with multimodal inputs. Based on the framework, we design and develop Smartboard, an agent-based interactive visualization system for fine-grained tactical analysis, especially for play design. Smartboard provides users with a structured process of setup, simulation, and evolution, allowing for iterative exploration of tactics based on specific personalized scenarios. We conduct case studies based on real-world basketball datasets to demonstrate the effectiveness and usefulness of our system.

The Impact of Default Mobile SDK Usage on Privacy and Data Protection

Are mobile app developers actively enabling data collection by advertisement and analytics companies, or are they unaware of the implications of using the provided software development kits (SDKs)? Given that the current mobile app ecosystem inadvertently involves collecting user data, which often infringes upon data protection and privacy standards, the question of the underlying reason for the permissibility of data processing arises. We contribute to this research for both Android and iOS by performing a two-step qualitative analysis. First, we conduct a structured documentation review of five advertisement and five analytics SDKs, focusing on privacy-related information. Subsequently, we implement a set of example apps utilizing the basic functionality of each SDK. This custom utilization of the SDK allows us to perform a fine-grained traffic analysis of each required step from initialization until utilization. Our results show that only little guidance on data protection compliance is provided. The observed network traffic shows that overall data collection by SDKs is similar between operating systems and only requires basic usage by the developer to trigger. We discover that with current SDKs, developers have minimal influence over the collected data, as merely using the basic functionality already results in data collection, with advertisement SDKs collecting more data than analytics SDKs. Overall, we explain the observed data protection infringement in ongoing mobile privacy research by documenting how developers must bear with opaque SDKs that lead to data collection simply due to usage.

The cognitive critical brain: modulation of criticality in perception-related cortical regions

The constantly evolving world necessitates a brain that can swiftly adapt and respond to rapid changes. The brain, conceptualized as a system performing cognitive functions through collective neural activity, has been shown to maintain a resting state characterized by near-critical neural dynamics, positioning it to effectively respond to external stimuli. However, how near-criticality is dynamically modulated during task performance remains insufficiently understood. In this study, we utilized the prototypical Ising Hamiltonian model to investigate the modulation of near-criticality in neural activity at the cortical subsystem level during perceptual tasks. Specifically, we simulated 2D-Ising models in silico using structural MRI data and empirically estimated the system's state in vivo using functional MRI data. We first replicated previous findings that the resting state is typically near-critical as captured by the Ising model. Importantly, we observed heterogeneous changes in criticality across cortical subsystems during a naturalistic movie-watching task, with visual and auditory regions fine-tuned closer to criticality. A more fine-grained analysis of the ventral temporal cortex during an object recognition task further revealed that only regions selectively responsive to a specific object category were tuned closer to criticality when processing that object category. In conclusion, our study provides empirical evidence from the domain of perception supporting the cognitive critical brain hypothesis that modulating the criticality of subsystems within the brain's hierarchical and modular organization may be a fundamental mechanism for achieving diverse cognitive functions.

Language barriers in emergency ambulance calls for cardiac arrest: Cases of missing vital information.

In medical emergencies, phoning the ambulance service constitutes a high-stakes interaction. Call-takers rely on callers to provide information about the patient so they can promptly recognise the medical problem and take swift action to remedy it. When a language barrier exists between the call-taker and caller, this can add a further challenge, given that third-party interpreters are rarely engaged, especially for time-critical conditions such as cardiac arrest. Research in cardiac arrest calls has found that language barrier calls experience longer delays to critical points such as recognition of cardiac arrest and commencement of resuscitation. This study aimed to understand, in the absence of interpreters, the interactional challenges that emerged in language barrier emergency calls, as parties worked to communicate the nature of the medical problem. Based on a critical conversation analysis approach, we conducted fine-grained analysis of interactions in audio recordings and transcripts of 33 language barrier calls from an Australian ambulance service in 2019. We found that call takers regularly failed to recognise that the patient had a cardiac arrest. Non-fluent-English callers often provided vital information about the patient, which could have led to cardiac arrest recognition by the call-taker, however such information was missed if it was delivered in an unsolicited or atypical way. Opportunities to recognise cardiac arrest were also missed when call-takers did not probe further after such information was provided or did not provide enough interactional space for callers to complete their turns. We found that the main reason for delays in recognising cardiac arrest was a lack of mutual understanding, which most of the time seemed to remain unbeknownst to participants. The study makes recommendations for emergency medical dispatch centres to cater for language barrier calls, with the goal of fostering a more inclusive prehospital care system and addressing health disparities for non-fluent-English speakers.

Interplay of Linguistic and Semiotic Choices in Shaping Meaning: A Critical Analysis of Political Cartoon in Pakistani English Newspaper

In today's world, where nations are interconnected through media, the analysis of media discourse from different perspectives on societal issues has gained much currency. The present study analyzes the satirical representation of a newspaper cartoon to find out how these media persons utilize cartoons and the accompanying linguistic choices to affect readers' opinions and understandings. Guided by qualitative research methodology, the data for this study is collected in the form of one cartoon selected through purposive sampling from a Pakistani English newspaper i.e., Dawn. Taking insights from Barthes's (1957) theory of semiotics and Fairclough’s (1995) Critical Discourse Analysis (CDA) model, the study utilizes an integrated framework for a fine-grain analysis of cartoon discourse. The intermingling of these two models of analysis has helped the researchers to highlight how newspaper cartoon discourse can be utilized to construct and highlight certain ideologies. The selected cartoon under the caption "OIC seeks global intervention to halt Gaza genocide," highlights the themes of victimization and powerlessness. By utilizing the signs of a man and a wooden log for the OIC (Organization of Islamic Cooperation) and Gaza, respectively, the cartoonist has evoked feelings of sympathy for the innocent killing of Palestinians. Hence, the study has found a logical connection between linguistic and semiotic systems to highlight the OIC's inefficacy and the world's indifference and hypocritical response towards the humanitarian crisis in Gaza.

A modular machine learning tool for holistic and fine-grained behavioral analysis.

Artificial intelligence techniques offer promising avenues for exploring human body features from videos, yet no freely accessible tool has reliably provided holistic and fine-grained behavioral analyses to date. To address this, we developed a machine learning tool based on a two-level approach: a first lower-level processing using computer vision for extracting fine-grained and comprehensive behavioral features such as skeleton or facial points, gaze, and action units; a second level of machine learning classification coupled with explainability providing modularity, to determine which behavioral features are triggered by specific environments. To validate our tool, we filmed 16 participants across six conditions, varying according to the presence of a person ("Pers"), a sound ("Snd"), or silence ("Rest"), and according to emotional levels using self-referential ("Self") and control ("Ctrl") stimuli. We demonstrated the effectiveness of our approach by extracting and correcting behavior from videos using two computer vision software (OpenPose and OpenFace) and by training two algorithms (XGBoost and long short-term memory [LSTM]) to differentiate between experimental conditions. High classification rates were achieved for "Pers" conditions versus "Snd" or "Rest" (AUC = 0.8-0.9), with explainability revealing actions units and gaze as key features. Additionally, moderate classification rates were attained for "Snd" versus "Rest" (AUC = 0.7), attributed to action units, limbs and head points, as well as for "Self" versus "Ctrl" (AUC = 0.7-0.8), due to facial points. These findings were consistent with a more conventional hypothesis-driven approach. Overall, our study suggests that our tool is well suited for holistic and fine-grained behavioral analysis and offers modularity for extension into more complex naturalistic environments.

A novel spatiotemporal prediction approach to fill air pollution data gaps using mobile sensors, machine learning and citizen science techniques

Particulate Matter (PM) air pollution poses significant threats to public health. We introduce a novel machine learning methodology to predict PM2.5 levels at 30 m long segments along the roads and at a temporal scale of 10 seconds. A hybrid dataset was curated from an intensive PM campaign in Selly Oak, Birmingham, UK, utilizing citizen scientists and low-cost instruments strategically placed in static and mobile settings. Spatially resolved proxy variables, meteorological parameters, and PM properties were integrated, enabling a fine-grained analysis of PM2.5. Calibration involved three approaches: Standard Random Forest Regression, Sensor Transferability and Road Transferability Evaluations. This methodology significantly increased spatial resolution beyond what is possible with regulatory monitoring, thereby improving exposure assessments. The findings underscore the importance of machine learning approaches and citizen science in advancing our understanding of PM pollution, with a small number of participants significantly enhancing local air quality assessment for thousands of residents.

Pheromone relay networks in the honeybee: messenger workers distribute the queen’s fertility signal throughout the hive

BackgroundThe harmonious operation of many insect societies depends upon colony-wide dissemination of a non-volatile pheromone produced by a single queen, which informs workers of her presence. This represents a major challenge in large colonies. Honeybee colonies, which can exceed 60,000 bees, are believed to solve this challenge using ‘messenger’ workers that actively relay the queen pheromone throughout the hive. However, little is known about the structure and effectiveness of the underlying relay network or the biology of messaging.ResultsHere, we combine automated tracking with modelling to address these outstanding questions. We find that both queen movement and worker messaging play fundamental roles in queen pheromone dissemination. Fine-grained analyses of worker behaviour confirmed the existence of active messaging, as physical contacts with the queen caused workers to move faster and straighter, thereby accelerating pheromone transmission. Finally, we show that messaging follows a stereotypical developmental trajectory, resulting in an age-dependent hierarchical relay network, with the most intense messaging observed between three and five days of age, when workers undergo a suite of physiological changes associated with queen rearing.ConclusionsThese results suggest that the individuals that contribute most to advertising the presence of the queen are also the ones that control queen production.