arXiv subjects

Armelle Brun

Publications and source records attributed to Armelle Brun.

16 recordsLinked to original sources

arXiv2026

CAARL: In-Context Learning for Interpretable Co-Evolving Time Series Forecasting

In this paper we investigate forecasting coevolving time series that feature intricate dependencies and nonstationary dynamics by using an LLM Large Language Models approach We propose a novel modeling approach named ContextAware ARLLM CAARL that provides an interpretable framework to decode the contextual dynamics influencing changes in coevolving series CAARL decomposes time series into autoregressive segments constructs a temporal dependency graph and serializes this graph into a narrative to allow processing by LLM This design yields a chainofthoughtlike reasoning path where intermediate steps capture contextual dynamics and guide forecasts in a transparent manner By linking prediction to explicit reasoning traces CAARL enhances interpretability while maintaining accuracy Experiments on realworld datasets validate its effectiveness positioning CAARL as a competitive and interpretable alternative to stateoftheart forecasting methods

cs.LG↗

arXiv2025

A Frugal Model for Accurate Early Student Failure Prediction

Predicting student success or failure is vital for timely interventions and personalized support. Early failure prediction is particularly crucial, yet limited data availability in the early stages poses challenges, one of the possible solutions is to make use of additional data from other contexts, however, this might lead to overconsumption with no guarantee of better results. To address this, we propose the Frugal Early Prediction (FEP) model, a new hybrid model that selectively incorporates additional data, promoting data frugality and efficient resource utilization. Experiments conducted on a public dataset from a VLE demonstrate FEP's effectiveness in reducing data usage, a primary goal of this research.Experiments showcase a remarkable 27% reduction in data consumption, compared to a systematic use of additional data, aligning with our commitment to data frugality and offering substantial benefits to educational institutions seeking efficient data consumption. Additionally, FEP also excels in enhancing prediction accuracy. Compared to traditional approaches, FEP achieves an average accuracy gain of 7.3%. This not only highlights the practicality and efficiency of FEP but also its superiority in performance, while respecting resource constraints, providing beneficial findings for educational institutions seeking data frugality.

cs.CY↗

arXiv2024

Combining Objective and Subjective Perspectives for Political News Understanding

Researchers and practitioners interested in computational politics rely on automatic content analysis tools to make sense of the large amount of political texts available on the Web. Such tools should provide objective and subjective aspects at different granularity levels to make the analyses useful in practice. Existing methods produce interesting insights for objective aspects, but are limited for subjective ones, are often limited to national contexts, and have limited explainability. We introduce a text analysis framework which integrates both perspectives and provides a fine-grained processing of subjective aspects. Information retrieval techniques and knowledge bases complement powerful natural language processing components to allow a flexible aggregation of results at different granularity levels. Importantly, the proposed bottom-up approach facilitates the explainability of the obtained results. We illustrate its functioning with insights on news outlets, political orientations, topics, individual entities, and demographic segments. The approach is instantiated on a large corpus of French news, but is designed to work seamlessly for other languages and countries.

cs.CL↗

arXiv2023

Do Similar Entities have Similar Embeddings?

Knowledge graph embedding models (KGEMs) developed for link prediction learn vector representations for entities in a knowledge graph, known as embeddings. A common tacit assumption is the KGE entity similarity assumption, which states that these KGEMs retain the graph's structure within their embedding space, \textit{i.e.}, position similar entities within the graph close to one another. This desirable property make KGEMs widely used in downstream tasks such as recommender systems or drug repurposing. Yet, the relation of entity similarity and similarity in the embedding space has rarely been formally evaluated. Typically, KGEMs are assessed based on their sole link prediction capabilities, using ranked-based metrics such as Hits@K or Mean Rank. This paper challenges the prevailing assumption that entity similarity in the graph is inherently mirrored in the embedding space. Therefore, we conduct extensive experiments to measure the capability of KGEMs to cluster similar entities together, and investigate the nature of the underlying factors. Moreover, we study if different KGEMs expose a different notion of similarity. Datasets, pre-trained embeddings and code are available at: https://github.com/nicolas-hbt/similar-embeddings/.

cs.AI↗

arXiv2023

All Polarized but Still Different: a Multi-factorial Metric to Discriminate between Polarization Behaviors on Social Media

Online polarization has attracted the attention of researchers for many years. Its effects on society are a cause for concern, and the design of personalized depolarization strategies appears to be a key solution. Such strategies should rely on a fine and accurate measurement, and a clear understanding of polarization behaviors. However, the literature still lacks ways to characterize them finely. We propose GRAIL, the first individual polarization metric, relying on multiple factors. GRAIL assesses these factors through entropy and is based on an adaptable Generalized Additive Model. We evaluate the proposed metric on a Twitter dataset related to the highly controversial debate about the COVID-19 vaccine. Experiments confirm the ability of GRAIL to discriminate between polarization behaviors. To go further, we provide a finer characterization and explanation of the identified behaviors through an innovative evaluation framework.

cs.SI↗

arXiv2023

Gaining a better understanding of online polarization by approaching it as a dynamic process

Polarization is often a clich{\'e}, its conceptualization remains approximate and no consensus has been reached so far. Often simply seen as an inevitable result of the use of social networks, polarization nevertheless remains a complex social phenomenon that must be placed in a wider context. To contribute to a better understanding of polarization, we approach it as an evolving process, drawing on a dual expertise in political and data sciences. We compare the polarization process between one mature debate (COVID-19 vaccine) and one emerging debate (Ukraine conflict) at the time of data collection. Both debates are studied on Twitter users, a highly politicized population, and on the French population to provide key elements beyond the traditional US context. This unprecedented analysis confirms that polarization varies over time, through a succession of specific periods, whose existence and duration depend on the maturity of the debate. Importantly, we highlight that polarization is paced by context-related events. Bearing this in mind, we pave the way for a new generation of personalized depolarization strategies, adapted to the context and maturity of debates.

cs.SI↗

arXiv2023

PyGraft: Configurable Generation of Synthetic Schemas and Knowledge Graphs at Your Fingertips

Knowledge graphs (KGs) have emerged as a prominent data representation and management paradigm. Being usually underpinned by a schema (e.g., an ontology), KGs capture not only factual information but also contextual knowledge. In some tasks, a few KGs established themselves as standard benchmarks. However, recent works outline that relying on a limited collection of datasets is not sufficient to assess the generalization capability of an approach. In some data-sensitive fields such as education or medicine, access to public datasets is even more limited. To remedy the aforementioned issues, we release PyGraft, a Python-based tool that generates highly customized, domain-agnostic schemas and KGs. The synthesized schemas encompass various RDFS and OWL constructs, while the synthesized KGs emulate the characteristics and scale of real-world KGs. Logical consistency of the generated resources is ultimately ensured by running a description logic (DL) reasoner. By providing a way of generating both a schema and KG in a single pipeline, PyGraft's aim is to empower the generation of a more diverse array of KGs for benchmarking novel approaches in areas such as graph-based machine learning (ML), or more generally KG processing. In graph-based ML in particular, this should foster a more holistic evaluation of model performance and generalization capability, thereby going beyond the limited collection of available benchmarks. PyGraft is available at: https://github.com/nicolas-hbt/pygraft.

cs.AI↗

arXiv2023

Schema First! Learn Versatile Knowledge Graph Embeddings by Capturing Semantics with MASCHInE

Knowledge graph embedding models (KGEMs) have gained considerable traction in recent years. These models learn a vector representation of knowledge graph entities and relations, a.k.a. knowledge graph embeddings (KGEs). Learning versatile KGEs is desirable as it makes them useful for a broad range of tasks. However, KGEMs are usually trained for a specific task, which makes their embeddings task-dependent. In parallel, the widespread assumption that KGEMs actually create a semantic representation of the underlying entities and relations (e.g., project similar entities closer than dissimilar ones) has been challenged. In this work, we design heuristics for generating protographs -- small, modified versions of a KG that leverage RDF/S information. The learnt protograph-based embeddings are meant to encapsulate the semantics of a KG, and can be leveraged in learning KGEs that, in turn, also better capture semantics. Extensive experiments on various evaluation benchmarks demonstrate the soundness of this approach, which we call Modular and Agnostic SCHema-based Integration of protograph Embeddings (MASCHInE). In particular, MASCHInE helps produce more versatile KGEs that yield substantially better performance for entity clustering and node classification tasks. For link prediction, using MASCHinE substantially increases the number of semantically valid predictions with equivalent rank-based performance.

cs.AI↗

arXiv2023

A Multi-Factorial Analysis of Polarization on Social Media

Polarization is an increasingly worrying phenomenon within social media. Recent work has made it possible to detect and even quantify polarization. Nevertheless, the few existing metrics, although defined in a continuous space, often lead to a unimodal distribution of data once applied to users' interactions, making the distinction between polarized and non-polarized users difficult to draw. Furthermore, each metric relies on a single factor and does not reflect the overall user behavior. Modeling polarization in a single form runs the risk of obscuring inter-individual differences. In this paper, we propose to have a deeper look at polarized online behaviors and to compare individual metrics. We collected about 300K retweets from 1K French users between January and July 2022 on Twitter. Each retweet is related to the highly controversial vaccine debate. Results show that a multi-factorial analysis leads to the identification of distinct and potentially explainable behavioral classes. This finer understanding of behaviors is an essential step to adapt news recommendation strategies so that no user gets locked into an echo chamber or filter bubble.

cs.SI↗

arXiv2023

Treat Different Negatives Differently: Enriching Loss Functions with Domain and Range Constraints for Link Prediction

Knowledge graph embedding models (KGEMs) are used for various tasks related to knowledge graphs (KGs), including link prediction. They are trained with loss functions that consider batches of true and false triples. However, different kinds of false triples exist and recent works suggest that they should not be valued equally, leading to specific negative sampling procedures. In line with this recent assumption, we posit that negative triples that are semantically valid w.r.t. signatures of relations (domain and range) are high-quality negatives. Hence, we enrich the three main loss functions for link prediction such that all kinds of negatives are sampled but treated differently based on their semantic validity. In an extensive and controlled experimental setting, we show that the proposed loss functions systematically provide satisfying results which demonstrates both the generality and superiority of our proposed approach. In fact, the proposed loss functions (1) lead to better MRR and Hits@10 values, and (2) drive KGEMs towards better semantic correctness as measured by the Sem@K metric. This highlights that relation signatures globally improve KGEMs, and thus should be incorporated into loss functions. Domains and ranges of relations being largely available in schema-defined KGs, this makes our approach both beneficial and widely usable in practice.

cs.LG↗

arXiv2023

Characterizing Financial Market Coverage using Artificial Intelligence

This paper scrutinizes a database of over 4900 YouTube videos to characterize financial market coverage. Financial market coverage generates a large number of videos. Therefore, watching these videos to derive actionable insights could be challenging and complex. In this paper, we leverage Whisper, a speech-to-text model from OpenAI, to generate a text corpus of market coverage videos from Bloomberg and Yahoo Finance. We employ natural language processing to extract insights regarding language use from the market coverage. Moreover, we examine the prominent presence of trending topics and their evolution over time, and the impacts that some individuals and organizations have on the financial market. Our characterization highlights the dynamics of the financial market coverage and provides valuable insights reflecting broad discussions regarding recent financial events and the world economy.

q-fin.ST↗

arXiv2023

Sem@$K$: Is my knowledge graph embedding model semantic-aware?

Using knowledge graph embedding models (KGEMs) is a popular approach for predicting links in knowledge graphs (KGs). Traditionally, the performance of KGEMs for link prediction is assessed using rank-based metrics, which evaluate their ability to give high scores to ground-truth entities. However, the literature claims that the KGEM evaluation procedure would benefit from adding supplementary dimensions to assess. That is why, in this paper, we extend our previously introduced metric Sem@K that measures the capability of models to predict valid entities w.r.t. domain and range constraints. In particular, we consider a broad range of KGs and take their respective characteristics into account to propose different versions of Sem@K. We also perform an extensive study to qualify the abilities of KGEMs as measured by our metric. Our experiments show that Sem@K provides a new perspective on KGEM quality. Its joint analysis with rank-based metrics offers different conclusions on the predictive power of models. Regarding Sem@K, some KGEMs are inherently better than others, but this semantic superiority is not indicative of their performance w.r.t. rank-based metrics. In this work, we generalize conclusions about the relative performance of KGEMs w.r.t. rank-based and semantic-oriented metrics at the level of families of models. The joint analysis of the aforementioned metrics gives more insight into the peculiarities of each model. This work paves the way for a more comprehensive evaluation of KGEM adequacy for specific downstream tasks.

cs.LG↗

arXiv2020

Multi-source Data Mining for e-Learning

Data mining is the task of discovering interesting, unexpected or valuable structures in large datasets and transforming them into an understandable structure for further use . Different approaches in the domain of data mining have been proposed, among which pattern mining is the most important one. Pattern mining mining involves extracting interesting frequent patterns from data. Pattern mining has grown to be a topic of high interest where it is used for different purposes, for example, recommendations. Some of the most common challenges in this domain include reducing the complexity of the process and avoiding the redundancy within the patterns. So far, pattern mining has mainly focused on the mining of a single data source. However, with the increase in the amount of data, in terms of volume, diversity of sources and nature of data, mining multi-source and heterogeneous data has become an emerging challenge in this domain. This challenge is the main focus of our work where we propose to mine multi-source data in order to extract interesting frequent patterns.

cs.DB↗

arXiv2019

Lead2Gold: Towards exploiting the full potential of noisy transcriptions for speech recognition

The transcriptions used to train an Automatic Speech Recognition (ASR) system may contain errors. Usually, either a quality control stage discards transcriptions with too many errors, or the noisy transcriptions are used as is. We introduce Lead2Gold, a method to train an ASR system that exploits the full potential of noisy transcriptions. Based on a noise model of transcription errors, Lead2Gold searches for better transcriptions of the training data with a beam search that takes this noise model into account. The beam search is differentiable and does not require a forced alignment step, thus the whole system is trained end-to-end. Lead2Gold can be viewed as a new loss function that can be used on top of any sequence-to-sequence deep neural network. We conduct proof-of-concept experiments on noisy transcriptions generated from letter corruptions with different noise levels. We show that Lead2Gold obtains a better ASR accuracy than a competitive baseline which does not account for the (artificially-introduced) transcription noise.

cs.CL↗

arXiv2019

Multi-source Relations for Contextual Data Mining in Learning Analytics

The goals of Learning Analytics (LA) are manifold, among which helping students to understand their academic progress and improving their learning process, which are at the core of our work. To reach this goal, LA relies on educational data: students' traces of activities on VLE, or academic, socio-demographic information, information about teachers, pedagogical resources, curricula, etc. The data sources that contain such information are multiple and diverse. Data mining, specifically pattern mining, aims at extracting valuable and understandable information from large datasets. In our work, we assume that multiple educational data sources form a rich dataset that can result in valuable patterns. Mining such data is thus a promising way to reach the goal of helping students. However, heterogeneity and interdependency within data lead to high computational complexity. We thus aim at designing low complex pattern mining algorithms that mine multi-source data, taking into consideration the dependency and heterogeneity among sources. The patterns formed are meaningful and interpretable, they can thus be directly used for students.

cs.DB↗

arXiv2019

Learning Analytics Made in France: The METALproject

This paper presents the METAL project, an ongoing French open Learning Analytics (LA) project for secondary school, that aims at improving the quality of the learning process. The originality of METAL is that it relies on research through exploratory activities and focuses on all the aspects of a Learning Analytics implementation. This large-scale project includes many concerns, divided into 4 main actions. (1) data management: multi-source data identification, collection and storage, selection and promotion of standards, and design and development of an open-source Learning Record Store (LRS); (2) data visualization: learner and teacher dashboards, with a design that relies on the co-conception with final users, including trust and usability concerns; (3) data exploitation: study of the link between gaze and memory of learners, design of explainable multi-source data-mining algorithms, including ethics and privacy concerns. An additional key of originality lies in the global dissemination of LA at an institution level or at a broader level such as a territory, at the opposite on many projects that focus on a specific school or a school curriculum. Each of these aspects is a hot topic in the literature. Taking into account all of them in a holistic view of education is an additional added value of the project.

cs.CY↗