arXiv subjects

Mariano G. Beiró

Publications and source records attributed to Mariano G. Beiró.

11 recordsLinked to original sources

arXiv2026

QuanReview: Offline, Auditable Reconciliation of Human and LLM Span Annotations

Structured span annotations, such as quantities with their units, uncertainty modifiers, and event classes, are expensive to create and hard to keep trustworthy once language models enter the loop. We present QuanReview, an open-source system for auditing and correcting such annotation layers. QuanReview aligns two annotation streams over the same documents at character level, resolves unambiguous cases by an explicit and logged policy, and routes candidate conflicts to a browser-based adjudication interface where reviewers accept either side, build field-level hybrids, or flag items for re-annotation. A campaign manager assigns documents to multiple annotators with configurable redundancy, computes agreement at document and span level, auto-merges unanimous documents, and exports the corrected layer in the original file format, so that it can replace the original annotation files directly. Applied to a 4,457-record humanitarian benchmark and an LLM extraction stream, the system fully auto-merged 8% of documents, applied automatic policy decisions to a further 1,513 records, and concentrated human attention on 3,131 candidate conflicts, a mean of 5.4 per reviewed document.

cs.CL↗

arXiv2026

Linnaeus: A Hierarchical, Multi-Label Framework for Autonomous System Classification

Autonomous systems (ASes) play diverse roles in today's Internet, from community and research backbones to hyperscale content providers and submarine-cable operators. However, existing taxonomies based solely on network-level features fail to capture their semantic and operational heterogeneity. In this paper, we present Linnaeus, a hierarchical AS-classification framework that combines network-centric data (e.g., topology, BGP announcements) with rich non-network features and leverages domain-adapted large language models alongside traditional machine-learning techniques. Linnaeus provides a two-level taxonomy with 18 top-level and 38 second-level classes, supports multi-label assignments to reflect hybrid roles (e.g., research backbone and transit provider), and provides an end-to-end pipeline from data ingestion to label inference. On a manually annotated dataset of nearly 2,000 ASes, Linnaeus achieves an overall precision and recall of 0.83 and 0.76, respectively. We further demonstrate its practical value through case studies, highlighting Linnaeus's potential to reveal both structural and semantic dimensions of Internet infrastructure.

cs.NI↗

arXiv2025

Characterizing User Behavior: The Interplay Between Mobility Patterns and Mobile Traffic

Mobile devices have become essential for capturing human activity, and eXtended Data Records (XDRs) offer rich opportunities for detailed user behavior modeling, which is useful for designing personalized digital services. Previous studies have primarily focused on aggregated mobile traffic and mobility analyses, often neglecting individual-level insights. This paper introduces a novel approach that explores the dependency between traffic and mobility behaviors at the user level. By analyzing 13 individual features that encompass traffic patterns and various mobility aspects, we enhance the understanding of how these behaviors interact. Our advanced user modeling framework integrates traffic and mobility behaviors over time, allowing for fine-grained dependencies while maintaining population heterogeneity through user-specific signatures. Furthermore, we develop a Markov model that infers traffic behavior from mobility and vice versa, prioritizing significant dependencies while addressing privacy concerns. Using a week-long XDR dataset from 1,337,719 users across several provinces in Chile, we validate our approach, demonstrating its robustness and applicability in accurately inferring user behavior and matching mobility and traffic profiles across diverse urban contexts.

cs.NI↗

arXiv2025

Beyond Sentiment: Examining the Role of Moral Foundations in User Engagement with News on Twitter

This study uses sentiment analysis and the Moral Foundations Theory (MFT) to characterise news content in social media and examine its association with user engagement. We employ Natural Language Processing to quantify the moral and affective linguistic markers. At the same time, we automatically define thematic macro areas of news from major U.S. news outlets and their Twitter followers (Jan 2020 - Mar 2021). By applying Non-Negative Matrix Factorisation to the obtained linguistic features we extract clusters of similar moral and affective profiles, and we identify the emotional and moral characteristics that mostly explain user engagement via regression modelling. We observe that Surprise, Trust, and Harm are crucial elements explaining user engagement and discussion length and that Twitter content from news media outlets has more explanatory power than their linked articles. We contribute with actionable findings evidencing the potential impact of employing specific moral and affective nuances in public and journalistic discourse in today's communication landscape. In particular, our results emphasise the need to balance engagement strategies with potential priming risks in our evolving media landscape.

cs.SI↗

arXiv2024

Signs of criticality in social explosions

The success of an on-line movement could be defined in terms of the shift to large-scale and the later off-line massive street actions of protests. The role of social media in this process is to facilitate the transformation from small or local feelings of disagreement into large-scale social actions. The way how social media achieves that effect is by growing clusters of people and groups with similar effervescent feelings, which in another case would never be in communication. It is natural to think that these kinds of macro social actions, as a consequence of the spontaneous and massive interactions, will attain the growth and divergence of the correlation length, giving rise to important simplifications on several statistics. In this work, we report the presence of signs of criticality in social demonstrations. Namely, the same power-law exponents are found whenever the distributions are calculated, either considering the same windows-time or the same number of hashtags. The exponents for the distributions during the event were found to be smaller than before (and after) the event. The latter also happens whenever the hashtags are counted only once per user or if all their usages are considered. By means of network representations, we show that the systems present two kinds of high correlations, characterised by either high or low values of modularity. The temporal points of high modularity are characterised by a sustained correlation while the ones of low modularity are characterised by a punctual correlation. The importance of analysing systems near a critical point is that any small disturbance can escalate and induce large-scale -- nationwide -- chain reactions.

physics.soc-ph↗

arXiv2021

Designing weighted and multiplex networks for deep learning user geolocation in Twitter

Predicting the geographical location of users of social media like Twitter has found several applications in health surveillance, emergency monitoring, content personalization, and social studies in general. In this work we contribute to the research in this area by designing and evaluating new methods based on the literature of weighted multigraphs combined with state-of-the-art deep learning techniques. The explored methods depart from a similar underlying structure (that of an extended mention and/or follower network) but use different information processing strategies, e.g., information diffusion through transductive and inductive algorithms -- RGCNs and GraphSAGE, respectively -- and node embeddings with Node2vec+. These graphs are then combined with attention mechanisms to incorporate the users' text view into the models. We assess the performance of each of these methods and compare them to baseline models in the publicly available Twitter-US dataset; we also make a new dataset available based on a large Twitter capture in Latin America. Finally, our work discusses the limitations and validity of the comparisons among methods in the context of different label definitions and metrics.

cs.SI↗

arXiv2020

Evolution of the political opinion landscape during electoral periods

We present a study of the evolution of the political landscape during the 2015 and 2019 presidential elections in Argentina, based on the data obtained from the micro-blogging platform Twitter. We build a semantic network based on the hashtags used by all the users following at least one of the main candidates. With this network we can detect the topics that are discussed in the society. At a difference with most studies of opinion on social media, we do not choose the topics a priori, they naturally emerge from the community structure of the semantic network instead. We assign to each user a dynamical topic vector which measures the evolution of her/his opinion in this space and allows us to monitor the similarities and differences among groups of supporters of different candidates. Our results show that the method is able to detect the dynamics of formation of opinion on different topics and, in particular, it can capture the reshaping of the political opinion landscape which has led to the inversion of result between the two rounds of the 2015 election.

cs.SI↗

arXiv2020

Learning language variations in news corpora through differential embeddings

There is an increasing interest in the NLP community in capturing variations in the usage of language, either through time (i.e., semantic drift), across regions (as dialects or variants) or in different social contexts (i.e., professional or media technolects). Several successful dynamical embeddings have been proposed that can track semantic change through time. Here we show that a model with a central word representation and a slice-dependent contribution can learn word embeddings from different corpora simultaneously. This model is based on a star-like representation of the slices. We apply it to The New York Times and The Guardian newspapers, and we show that it can capture both temporal dynamics in the yearly slices of each corpus, and language variations between US and UK English in a curated multi-source corpus. We provide an extensive evaluation of this methodology.

cs.CL↗

arXiv2018

Shopping Mall Attraction and Social Mixing at a City Scale

The social inclusion aspects of shopping malls and their effects on our understanding of urban spaces have been a controversial argument largely discussed in the literature. Shopping malls offer an open, safe and democratic version of the public space. Many of their detractors suggest that malls target their customers in subtle ways, promoting social exclusion. In this work, we analyze whether malls offer opportunities for social mixing by analyzing the patterns of shopping mall visits in a large Latin-American city: Santiago de Chile. We use a large XDR (Data Detail Records) dataset from a telecommunication company to analyze the mobility of $387,152$ cell phones around $16$ large malls in Santiago de Chile during one month. We model the influx of people to malls in terms of a gravity model of mobility, and we are able to predict the customer profile distribution of each mall, explaining it in terms of mall location, the population distribution, and mall size. Then, we analyze the concept of social attraction, expressed as people from low and middle classes being attracted by malls that target high-income customers. We include a social attraction factor in our model and find that it is negligible in the process of choosing a mall. We observe that social mixing arises only in peripheral malls located farthest from the city center, which both low and middle class people visit. Using a co-visitation model we show that people tend to choose a restricted profile of malls according to their socio-economic status and their distance from the mall. We conclude that the potential for social mixing in malls could be capitalized by designing public policies regarding transportation and mobility.

cs.SI↗

arXiv2015

Router-level community structure of the Internet Autonomous Systems

The Internet is composed of routing devices connected between them and organized into independent administrative entities: the Autonomous Systems. The existence of different types of Autonomous Systems (like large connectivity providers, Internet Service Providers or universities) together with geographical and economical constraints, turns the Internet into a complex modular and hierarchical network. This organization is reflected in many properties of the Internet topology, like its high degree of clustering and its robustness. In this work, we study the modular structure of the Internet router-level graph in order to assess to what extent the Autonomous Systems satisfy some of the known notions of community structure. We show that the modular structure of the Internet is much richer than what can be captured by the current community detection methods, which are severely affected by resolution limits and by the heterogeneity of the Autonomous Systems. Here we overcome this issue by using a multiresolution detection algorithm combined with a small sample of nodes. We also discuss recent work on community structure in the light of our results.

cs.NI↗

arXiv2012

Obtaining Communities with a Fitness Growth Process

The study of community structure has been a hot topic of research over the last years. But, while successfully applied in several areas, the concept lacks of a general and precise notion. Facts like the hierarchical structure and heterogeneity of complex networks make it difficult to unify the idea of community and its evaluation. The global functional known as modularity is probably the most used technique in this area. Nevertheless, its limits have been deeply studied. Local techniques as the ones by Lancichinetti et al. and Palla et al. arose as an answer to the resolution limit and degeneracies that modularity has. Here we start from the algorithm by Lancichinetti et al. and propose a unique growth process for a fitness function that, while being local, finds a community partition that covers the whole network, updating the scale parameter dynamically. We test the quality of our results by using a set of benchmarks of heterogeneous graphs. We discuss alternative measures for evaluating the community structure and, in the light of them, infer possible explanations for the better performance of local methods compared to global ones in these cases.

physics.soc-ph↗