Josiane Mothe

Eduard Poesina, Radu Tudor Ionescu, Josiane Mothe

To date, query performance prediction (QPP) in the context of content-based image retrieval remains a largely unexplored task, especially in the query-by-example scenario, where the query is an image. To boost the exploration of the QPP task in image retrieval, we propose the first benchmark for image query performance prediction (iQPP). First, we establish a set of four data sets (PASCAL VOC 2012, Caltech-101, ROxford5k and RParis6k) and estimate the ground-truth difficulty of each query as the average precision or the precision@k, using two state-of-the-art image retrieval models. Next, we propose and evaluate novel pre-retrieval and post-retrieval query performance predictors, comparing them with existing or adapted (from text to image) predictors. The empirical results show that most predictors do not generalize across evaluation scenarios. Our comprehensive experiments indicate that iQPP is a challenging benchmark, revealing an important research gap that needs to be addressed in future work. We release our code and data as open source at https://github.com/Eduard6421/iQPP, to foster future research.

Josiane Mothe

Isidore of Seville is credited with the adage that it is language that gives birth to a people, and not the other way around , underlining the profound role played by language in the formation of cultural and social identity. Today, of the more than 7100 languages listed, a significant number are endangered. Since the 1970s, linguists, information seekers and enthusiasts have helped develop digital resources and automatic tools to support a wide range of languages, including endangered ones. The advent of Large Language Model (LLM) technologies holds both promise and peril. They offer unprecedented possibilities for the translation and generation of content and resources, key elements in the preservation and revitalisation of languages. They also present threat of homogenisation, cultural oversimplification and the further marginalisation of already vulnerable languages. The talk this paper is based on has proposed an initiatory journey, exploring the potential paths and partnerships between technology and tradition, with a particular focus on the Occitan language. Occitan is a language from Southern France, parts of Spain and Italy that played a major cultural and economic role, particularly in the Middle Ages. It is now endangered according to UNESCO. The talk critically has examined how human expertise and artificial intelligence can work together to offer hope for preserving the linguistic diversity that forms the foundation of our global and especially our European heritage while addressing some of the ethical and practical challenges that accompany the use of these powerful technologies. This paper is based on the keynote I gave at the 46th European Conference on Information Retrieval (ECIR 2024). As an alternative to reading this paper, a video talk is available online. 1 Date: 26 March 2024.

Tilahun Yeshambel, Moncef Garouani, Josiane Mothe

Neural retrieval and GPT-style generative models rely on large, high-quality supervised data, which is still scarce for low-resource languages such as Amharic. We release an Amharic data resource consisting of two datasets that supports research on (i) neural retrieval-ranking and (ii) instruction-following text generation. The retrieval-ranking dataset contains 1,091 manually verified query-positive-negative document triplets drawn from diverse Amharic sources and constructed to support contrastive training and benchmarking of neural retrievers (e.g., DPR, ColBERT-style late interaction and SPLADE-style sparse neural retrieval). Triplets are created through a combination of expert-curated queries, web-derived queries, and LLM-assisted generation, with positive/negative documents selected from the web or synthesized by LLMs and then validated by native speakers. The instruction prompt-response dataset comprises 6,285 Amharic prompt-response pairs spanning multiple domains and instruction types, generated with several LLMs and refined through manual review and correction for grammaticality, relevance, fluency, and factual plausibility. We release both datasets with standardized splits and formats (CSV,JSON,JSONL) to enable reproducible work on Amharic retrieval, ranking, and generative modelling. These datasets also come with a methodology that can be generalized to other low-resource languages.

Moncef Garouani, Josiane Mothe

Neural retrievers are effective but brittle: underspecified or ambiguous queries can misdirect ranking even when relevant documents exist. Existing approaches address this brittleness only partially: LLMs rewrite queries without retriever feedback, and explainability methods identify misleading tokens but are used for post-hoc analysis. We close this loop and propose an attribution-guided query rewriting method that uses token-level explanations to guide query rewriting. For each query, we compute gradient-based token attributions from the retriever and then use these scores as soft guidance in a structured prompt to an LLM that clarifies weak or misleading query components while preserving intent. Evaluated on BEIR collections, the resulting rewrites consistently improve retrieval effectiveness over strong baselines, with larger gains for implicit or ambiguous information needs.

Victor Morand, Josiane Mothe, Benjamin Piwowarski

Named entities are fundamental building blocks of knowledge in text, grounding factual information and structuring relationships within language. Despite their importance, it remains unclear how Large Language Models (LLMs) internally represent entities. Prior research has primarily examined explicit relationships, but little is known about entity representations themselves. We introduce entity mention reconstruction as a novel framework for studying how LLMs encode and manipulate entities. We investigate whether entity mentions can be generated from internal representations, how multi-token entities are encoded beyond last-token embeddings, and whether these representations capture relational knowledge. Our proposed method, leveraging _task vectors_, allows to consistently generate multi-token mentions from various entity representations derived from the LLMs hidden states. We thus introduce the _Entity Lens_, extending the _logit-lens_ to predict multi-token mentions. Our results bring new evidence that LLMs develop entity-specific mechanisms to represent and manipulate any multi-token entities, including those unseen during training. Our code is avalable at https://github.com/VictorMorand/EntityRepresentations .

Hugo Carlesso, Josiane Mothe, Radu Tudor Ionescu

Hyperspectral imaging (HSI) captures detailed spectral signatures across hundreds of contiguous bands per pixel, being indispensable for remote sensing applications such as land-cover classification, change detection, and environmental monitoring. Due to the high dimensionality of HSI data and the slow rate of data transfer in satellite-based systems, compact and efficient models are required to support onboard processing and minimize the transmission of redundant or low-value data, e.g. cloud-covered areas. To this end, we introduce a novel curriculum multi-task self-supervised learning (CMTSSL) framework designed for lightweight architectures for HSI analysis. CMTSSL integrates masked image modeling with decoupled spatial and spectral jigsaw puzzle solving, guided by a curriculum learning strategy that progressively increases data complexity during self-supervision. This enables the encoder to jointly capture fine-grained spectral continuity, spatial structure, and global semantic features. Unlike prior dual-task SSL methods, CMTSSL simultaneously addresses spatial and spectral reasoning within a unified and computationally efficient design, being particularly suitable for training lightweight models for onboard satellite deployment. We validate our approach on four public benchmark datasets, demonstrating consistent gains in downstream segmentation tasks, using architectures that are over 16,000x lighter than some state-of-the-art models. These results highlight the potential of CMTSSL in generalizable representation learning with lightweight architectures for real-world HSI applications. Our code is publicly available at https://github.com/hugocarlesso/CMTSSL.

Hugo Carlesso, Maria Eliza Patulea, Moncef Garouani et al.

Mixup has become a popular augmentation strategy for image classification, yet its naive pixel-wise interpolation often produces unrealistic images that can hinder learning, particularly in high-stakes medical applications. We propose GeMix, a two-stage framework that replaces heuristic blending with a learned, label-aware interpolation powered by class-conditional GANs. First, a StyleGAN2-ADA generator is trained on the target dataset. During augmentation, we sample two label vectors from Dirichlet priors biased toward different classes and blend them via a Beta-distributed coefficient. Then, we condition the generator on this soft label to synthesize visually coherent images that lie along a continuous class manifold. We benchmark GeMix on the large-scale COVIDx-CT-3 dataset using three backbones (ResNet-50, ResNet-101, EfficientNet-B0). When combined with real data, our method increases macro-F1 over traditional mixup for all backbones, reducing the false negative rate for COVID-19 detection. GeMix is thus a drop-in replacement for pixel-space mixup, delivering stronger regularization and greater semantic fidelity, without disrupting existing training pipelines. We publicly release our code at https://github.com/hugocarlesso/GeMix to foster reproducibility and further research.

Eduard Poesina, Adriana Valentina Costache, Adrian-Gabriel Chifu et al.

Text-to-image generation has recently emerged as a viable alternative to text-to-image retrieval, driven by the visually impressive results of generative diffusion models. Although query performance prediction is an active research topic in information retrieval, to the best of our knowledge, there is no prior study that analyzes the difficulty of queries (referred to as prompts) in text-to-image generation, based on human judgments. To this end, we introduce the first dataset of prompts which are manually annotated in terms of image generation performance. Additionally, we extend these evaluations to text-to-image retrieval by collecting manual annotations that represent retrieval performance. We thus establish the first joint benchmark for prompt and query performance prediction (PQPP) across both tasks, comprising over 10K queries. Our benchmark enables (i) the comparative assessment of prompt/query difficulty in both image generation and image retrieval, and (ii) the evaluation of prompt/query performance predictors addressing both generation and retrieval. We evaluate several pre- and post-generation/retrieval performance predictors, thus providing competitive baselines for future research. Our benchmark and code are publicly available at https://github.com/Eduard6421/PQPP.

Moncef Garouani, Josiane Mothe, Ayah Barhrhouj et al.

The rapid evolution of machine learning (ML) has led to the widespread adoption of complex "black box" models, such as deep neural networks and ensemble methods. These models exhibit exceptional predictive performance, making them invaluable for critical decision-making across diverse domains within society. However, their inherently opaque nature raises concerns about transparency and interpretability, making them untrustworthy decision support systems. To alleviate such a barrier to high-stakes adoption, research community focus has been on developing methods to explain black box models as a means to address the challenges they pose. Efforts are focused on explaining these models instead of developing ones that are inherently interpretable. Designing inherently interpretable models from the outset, however, can pave the path towards responsible and beneficial applications in the field of ML. In this position paper, we clarify the chasm between explaining black boxes and adopting inherently interpretable models. We emphasize the imperative need for model interpretability and, following the purpose of attaining better (i.e., more effective or efficient w.r.t. predictive performance) and trustworthy predictors, provide an experimental evaluation of latest hybrid learning methods that integrates symbolic knowledge into neural network predictors. We demonstrate how interpretable hybrid models could potentially supplant black box ones in different domains.

Nathalie Neptune, Josiane Mothe

The Amazon rain forest is a vital ecosystem that plays a crucial role in regulating the Earth's climate and providing habitat for countless species. Deforestation in the Amazon is a major concern as it has a significant impact on global carbon emissions and biodiversity. In this paper, we present a method for detecting deforestation in the Amazon using image pairs from Earth observation satellites. Our method leverages deep learning techniques to compare the images of the same area at different dates and identify changes in the forest cover. We also propose a visual semantic model that automatically annotates the detected changes with relevant keywords. The candidate annotation for images are extracted from scientific documents related to the Amazon region. We evaluate our approach on a dataset of Amazon image pairs and demonstrate its effectiveness in detecting deforestation and generating relevant annotations. Our method provides a useful tool for monitoring and studying the impact of deforestation in the Amazon. While we focus on environment applications of our work by using images of deforestation in the Amazon rain forest to demonstrate the effectiveness of our proposed approach, it is generic enough to be applied to other domains.

Hajar Dekdegue, Moncef Garouani, Josiane Mothe et al.

Interpreting the decision-making process of deep convolutional neural networks remains a central challenge in achieving trustworthy and transparent artificial intelligence. Explainable AI (XAI) techniques, particularly Class Activation Map (CAM) methods, are widely adopted to visualize the input regions influencing model predictions. Gradient-based approaches (e.g. Grad-CAM) provide highly discriminative, fine-grained details by computing gradients of class activations but often yield noisy and incomplete maps that emphasize only the most salient regions rather than the complete objects. Region-based approaches (e.g. Score-CAM) aggregate information over larger areas, capturing broader object coverage at the cost of over-smoothing and reduced sensitivity to subtle features. We introduce Fusion-CAM, a novel framework that bridges this explanatory gap by unifying both paradigms through a dedicated fusion mechanism to produce robust and highly discriminative visual explanations. Our method first denoises gradient-based maps, yielding cleaner and more focused activations. It then combines the refined gradient map with region-based maps using contribution weights to enhance class coverage. Finally, we propose an adaptive similarity-based pixel-level fusion that evaluates the agreement between both paradigms and dynamically adjusts the fusion strength. This adaptive mechanism reinforces consistent activations while softly blending conflicting regions, resulting in richer, context-aware, and input-adaptive visual explanations. Extensive experiments on standard benchmarks show that Fusion-CAM consistently outperforms existing CAM variants in both qualitative visualization and quantitative evaluation, providing a robust and flexible tool for interpreting deep neural networks.

Victor Morand, Nadi Tomeh, Josiane Mothe et al.

Identifying which text spans refer to entities -- mention detection -- is both foundational for information extraction and a known performance bottleneck. We introduce ToMMeR, a lightweight model (<300K parameters) probing mention detection capabilities from early LLM layers. Across 13 NER benchmarks, ToMMeR achieves 93\% recall zero-shot, with over 90\% precision using an LLM as a judge showing that ToMMeR rarely produces spurious predictions despite high recall. Cross-model analysis reveals that diverse architectures (14M-15B parameters) converge on similar mention boundaries (DICE >75\%), confirming that mention detection emerges naturally from language modeling. When extended with span classification heads, ToMMeR achieves near SOTA NER performance (80-87\% F1 on standard benchmarks). Our work provides evidence that structured entity representations exist in early transformer layers and can be efficiently recovered with minimal parameters.

Mohammad El Sakka, Caroline De Pourtales, Lotfi Chaari et al.

Remote sensing has emerged as a critical tool for large-scale Earth monitoring and land management. In this paper, we introduce AgriPotential, a novel benchmark dataset composed of Sentinel-2 satellite imagery spanning multiple months. The dataset provides pixel-level annotations of agricultural potentials for three major crop types - viticulture, market gardening, and field crops - across five ordinal classes. AgriPotential supports a broad range of machine learning tasks, including ordinal regression, multi-label classification, and spatio-temporal modeling. The data covers diverse areas in Southern France, offering rich spectral information. AgriPotential is the first public dataset designed specifically for agricultural potential prediction, aiming to improve data-driven approaches to sustainable land use planning. The dataset and the code are freely accessible at: https://zenodo.org/records/15556484

Adrian-Gabriel Chifu, Sébastien Déjean, Josiane Mothe et al.

Query Performance Prediction (QPP) estimates retrieval systems effectiveness for a given query, offering valuable insights for search effectiveness and query processing. Despite extensive research, QPPs face critical challenges in generalizing across diverse retrieval paradigms and collections. This paper provides a comprehensive evaluation of state-of-the-art QPPs (e.g. NQC, UQC), LETOR-based features, and newly explored dense-based predictors. Using diverse sparse rankers (BM25, DFree without and with query expansion) and hybrid or dense (SPLADE and ColBert) rankers and diverse test collections ROBUST, GOV2, WT10G, and MS MARCO; we investigate the relationships between predicted and actual performance, with a focus on generalization and robustness. Results show significant variability in predictors accuracy, with collections as the main factor and rankers next. Some sparse predictors perform somehow on some collections (TREC ROBUST and GOV2) but do not generalise to other collections (WT10G and MS-MARCO). While some predictors show promise in specific scenarios, their overall limitations constrain their utility for applications. We show that QPP-driven selective query processing offers only marginal gains, emphasizing the need for improved predictors that generalize across collections, align with dense retrieval architectures and are useful for downstream applications.

Michel Boeglin, David Kahn, Josiane Mothe et al.

D4R is a digital platform designed to assist non-technical users, particularly historians, in exploring textual documents through advanced graphical tools for text analysis and knowledge extraction. By leveraging a large language model, D4R translates natural language questions into Cypher queries, enabling the retrieval of data from a Neo4J database. A user-friendly graphical interface allows for intuitive interaction, enabling users to navigate and analyse complex relational data extracted from unstructured textual documents. Originally designed to bridge the gap between AI technologies and historical research, D4R's capabilities extend to various other domains. A demonstration video and a live software demo are available.

Tilahun Yeshambel, Moncef Garouani, Serge Molina et al.

This paper reports some difficulties and some results when using dense retrievers on Amharic, one of the low-resource languages spoken by 120 millions populations. The efforts put and difficulties faced by University Addis Ababa toward Amharic Information Retrieval will be developed during the presentation.

Hèctor Loopez Hidalgo, Michel Boeglin, David Kahn et al.

Semantic relations among entities are a widely accepted method for relation extraction. PromptORE (Prompt-based Open Relation Extraction) was designed to improve relation extraction with Large Language Models on generalistic documents. However, it is less effective when applied to historical documents, in languages other than English. In this study, we introduce an adaptation of PromptORE to extract relations from specialized documents, namely digital transcripts of trials from the Spanish Inquisition. Our approach involves fine-tuning transformer models with their pretraining objective on the data they will perform inference. We refer to this process as "biasing". Our Biased PromptORE addresses complex entity placements and genderism that occur in Spanish texts. We solve these issues by prompt engineering. We evaluate our method using Encoder-like models, corroborating our findings with experts' assessments. Additionally, we evaluate the performance using a binomial classification benchmark. Our results show a substantial improvement in accuracy -up to a 50% improvement with our Biased PromptORE models in comparison to the baseline models using standard PromptORE.

Md Siddiqur Rahman, Laurent Lapasset, Josiane Mothe

An aircraft conflict occurs when two or more aircraft cross at a certain distance at the same time. Specific air traffic controllers are assigned to solve such conflicts. A controller needs to consider various types of information in order to solve a conflict. The most common and preliminary information is the coordinate position of the involved aircraft. Additionally, a controller has to take into account more information such as flight planning, weather, restricted territory, etc. The most important challenges a controller has to face are: to think about the issues involved and make a decision in a very short time. Due to the increased number of aircraft, it is crucial to reduce the workload of the controllers and help them make quick decisions. A conflict can be solved in many ways, therefore, we consider this problem as a multi-label classification problem. In doing so, we are proposing a multi-label classification model which provides multiple heading advisories for a given conflict. This model we named CRMLnet is based on a novel application of a multi-layer neural network and helps the controllers in their decisions. When compared to other machine learning models, our CRMLnet has achieved the best results with an accuracy of 98.72% and ROC of 0.999. The simulated data set that we have developed and used in our experiments will be delivered to the research community.

Ismat Ara Reshma, Sylvain Cussat-Blanc, Radu Tudor Ionescu et al.

The class distribution of data is one of the factors that regulates the performance of machine learning models. However, investigations on the impact of different distributions available in the literature are very few, sometimes absent for domain-specific tasks. In this paper, we analyze the impact of natural and balanced distributions of the training set in deep learning (DL) models applied on histological images, also known as whole slide images (WSIs). WSIs are considered as the gold standard for cancer diagnosis. In recent years, researchers have turned their attention to DL models to automate and accelerate the diagnosis process. In the training of such DL models, filtering out the non-regions-of-interest from the WSIs and adopting an artificial distribution (usually, a balanced distribution) is a common trend. In our analysis, we show that keeping the WSIs data in their usual distribution (which we call natural distribution) for DL training produces fewer false positives (FPs) with comparable false negatives (FNs) than the artificially-obtained balanced distribution. We conduct an empirical comparative study with 10 random folds for each distribution, comparing the resulting average performance levels in terms of five different evaluation metrics. Experimental results show the effectiveness of the natural distribution over the balanced one across all the evaluation metrics.

Bernard Dousset, Josiane Mothe

COVID-19 is one of the most important topic these days, specifically on search engines and news. While fake news are easily shared, scientific papers are reliable sources where information can be extracted. With about 24,000 scientific publications on COVID-19 and related research on PUBMED, automatic computer-assisted analysis is required. In this paper, we develop two methodologies to get insights on specific sub-topics of interest and latest research sub-topics. They rely on natural language processing and graph-based visualizations. We run these methodologies on two cases: the virus origin and the uses of existing drugs.

Sébastien Déjean, Radu Tudor Ionescu, Josiane Mothe et al.

The goal of query performance prediction (QPP) is to automatically estimate the effectiveness of a search result for any given query, without relevance judgements. Post-retrieval features have been shown to be more effective for this task while being more expensive to compute than pre-retrieval features. Combining multiple post-retrieval features is even more effective, but state-of-the-art QPP methods are impossible to interpret because of the black-box nature of the employed machine learning models. However, interpretation is useful for understanding the predictive model and providing more answers about its behavior. Moreover, combining many post-retrieval features is not applicable to real-world cases, since the query running time is of utter importance. In this paper, we investigate a new framework for feature selection in which the trained model explains well the prediction. We introduce a step-wise (forward and backward) model selection approach where different subsets of query features are used to fit different models from which the system selects the best one. We evaluate our approach on four TREC collections using standard QPP features. We also develop two QPP features to address the issue of query-drift in the query feedback setting. We found that: (1) our model based on a limited number of selected features is as good as more complex models for QPP and better than non-selective models; (2) our model is more efficient than complex models during inference time since it requires fewer features; (3) the predictive model is readable and understandable; and (4) one of our new QPP features is consistently selected across different collections, proving its usefulness.

Léa Laporte, Rémi Flamary, Stephane Canu et al.

Feature selection in learning to rank has recently emerged as a crucial issue. Whereas several preprocessing approaches have been proposed, only a few works have been focused on integrating the feature selection into the learning process. In this work, we propose a general framework for feature selection in learning to rank using SVM with a sparse regularization term. We investigate both classical convex regularizations such as $\ell\_1$ or weighted $\ell\_1$ and non-convex regularization terms such as log penalty, Minimax Concave Penalty (MCP) or $\ell\_p$ pseudo norm with $p\textless{}1$. Two algorithms are proposed, first an accelerated proximal approach for solving the convex problems, second a reweighted $\ell\_1$ scheme to address the non-convex regularizations. We conduct intensive experiments on nine datasets from Letor 3.0 and Letor 4.0 corpora. Numerical results show that the use of non-convex regularizations we propose leads to more sparsity in the resulting models while prediction performance is preserved. The number of features is decreased by up to a factor of six compared to the $\ell\_1$ regularization. In addition, the software is publicly available on the web.