2024Activity reportProject-TeamCEDAR

7.1.1 ConnectionLens

• Name:
  Integration of heterogeneous data using information extraction
• Keyword:
  Data analysis
• Functional Description:
  ConnectionLens treats a set of heterogeneous, independently authored data sources as a single virtual graph, whereas nodes represent fine-granularity data items (relational tuples, attributes, key-value pairs, RDF, JSON or XML nodes…) and edges correspond either to structural connections (e.g., a tuple is in a database, an attribute is in a tuple, a JSON node has a parent…) or to similarity (sameAs) links. To further enrich the content journalists work with, we also apply entity extraction which enables to detect the people, organizations etc. mentioned in text, whether full-text or text snippets found e.g. in RDF or XML. ConnectionLens is thus capable of finding and exploiting connections present across heterogeneous data sources without requiring the user to specify any join predicate.
• URL:
  https://­team.­inria.­fr/­cedar/­connectionlens/
• Publications:
  hal-02934277, hal-02904797, hal-01841009
• Contact:
  Manolescu Ioana

7.1.2 Abstra

• Name:
  Abstra: Toward Generic Abstractions for Data of Any Model
• Keywords:
  Heterogeneous Data, Data Exploration, Data analysis, Databases, LOD - Linked open data
• Functional Description:
  Abstra computes a description meant for humans, based on the idea that, regardless of the syntax or the data model, any dataset holds some collections of entities/records, that are possibly linked with relationships. Abstra relies on a common graph representation of any incoming dataset, it leverages Information Extraction to detect what the dataset is about, and relies on an original algorithm for selecting the core entity collections and their relations. Abstractions are shown both as HTML text and a lightweight Entity-Relationship diagram. A GUI also allows to tune the abstraction parameters and explore the dataset.
• URL:
  https://­team.­inria.­fr/­cedar/­projects/­abstra/
• Contact:
  Ioana Manolescu

7.1.3 StatCheck

• Name:
  Fact-checking Multidimensional Statistic Claims in French
• Keywords:
  Machine learning, Databases, Natural language processing, Software engineering
• Scientific Description:
  To strengthen public trust and counter disinformation, computational fact-checking, leveraging digital data sources, attracts interest from the journalists and the computer science community. A particular class of interesting data sources comprises statistics, that is, numerical data compiled mostly by governments, administrations, and international organizations. Statistics are often multidimensional datasets, where multiple dimensions characterize one value, and the dimensions may be organized in hierarchies. This paper describes STATCHECK, a statistic fact-checking system jointly developed by the authors, which are either computer science researchers or fact-checking journalists working for a French-language media with a daily audience of more than 15 millions (aud, 2022). The technical novelty of STATCHECK is twofold: (i) we focus on multidimensional, complex-structure statistics, which have received little attention so far, despite their practical importance, and (ii) novel statistical claim extraction modules for French, an area where few resources exist. We validate the efficiency and quality of our system on large statistic datasets (hundreds of millions of facts), including the complete INSEE (French) and Eurostat (European Union) datasets, as well as French presidential election debates.
• Functional Description:
  StatCheck firstly allows the collection of data for its operation. Two types of data are collected: statistical tables and posts from social networks: - Acquisition of statistical files on the site of referent organisations (INSEE, Eurostat) - Extraction of statistical tables from these files, and storage of the extracted tables - Acquisition of political tweets from a list of accounts The application allows the detection, extraction and search of statistical facts: - Detection and extraction of statistical facts from Twitter posts (e.g. "Unemployment rate increased by 30% in 2023) - Search for statistical facts in our database. Display of the twenty most relevant statistical tables for a statistical fact - Automatic transcription of audio files to detect and extract transcripts of statistical facts.
• Release Contributions:
  - Redesign of the user interface - Modification of the software architecture - Addition of audio transcription
• URL:
  https://­cedar-rf.­saclay.­inria.­fr/
• Publications:
  hal-01496700, hal-01745768, hal-02121389, hal-01915148, hal-03767992, hal-03791175
• Contact:
  Ioana Manolescu
• Participants:
  Tien Duc Cao, Ioana Manolescu, Xavier Tannier, Oana-Denisa Balalau, Simon Ebel, Theo Galizzi

7.1.4 ConnectionStudio

• Keywords:
  Heterogeneous Data, Data Exploration
• Functional Description:

  ConnectionStudio integrates highly heterogeneous data into graphs, enriched with extracted entities. Studio users can discover the entities in their data, navigate across connections between datasets, explore and query the data in many ways. The Studio currently supports: CSV, JSON, XML, RDF, text, property graphs, all Office formats, and PDF datasets.

  ConnectionStudio is a novel front-end to ConnectionLens, Abstra and PathWays (see also the respective Web sites). Its own novel features are outlined in a CoopIS 2023 article.

• URL:
  https://­connectionstudio.­inria.­fr/
• Contact:
  Ioana Manolescu

7.1.5 FactSpotter

• Keywords:
  Factual Faithfulness, Text generation
• Functional Description:
  We propose a new metric that correctly identifies factual faithfulness, i.e., given a triple (subject, predicate, object), it decides if the triple is present in a generated text. We show that our metric FactSpotter achieves the highest correlation with human annotations on data correct- ness, data coverage, and relevance. In addition, FactSpotter can be used as a plug-in feature to improve the factual faithfulness of existing models.
• Contact:
  Kun Zhang
• Partner:
  Ecole Polytechnique

7.1.6 PathWays

• Name:
  PathWays: finding entity paths in heterogeneous data graphs
• Keywords:
  Named entities, Data Journalism, Heterogeneous Data
• Functional Description:
  PathWays models heteroegenous datasets in a graph (see ConnectionLens). To identify interesting paths in this graph, Pathways works on its (smaller) summary (see Abstra) for efficiency and optimisation. Then, it sorts paths by their potential interest (metric based on the entity found and the information diluation along the path) before evaluating them with the help of a new multi-query optimisation algorithm. Finally, PathWays shows the most interesting (evaluated) paths in the form of tables, wich are very easy to understanf for journalists who are at the initiative of this scenario.
• URL:
  https://­team.­inria.­fr/­cedar/­projects/­pathways/
• Contact:
  Ioana Manolescu

7.1.7 OpenIEEntity

• Name:
  Open Information Extraction with Entity Focused Constraints
• Keyword:
  Information extraction
• Functional Description:
  This tool takes in input a sentence and outputs the facts contained in the sentence, in the format (subject,predicate,object).
• Contact:
  Oana-Denisa Balalau

7.1.8 FactCheckBureau

• Name:
  FactCheckBureau: Build Your Own Fact-Check Analysis Pipeline
• Keywords:
  Fact Check Retireval, Fact-checking
• Functional Description:
  FactCheckBurea is an end-to-end solution that enables researchers to easily and interactively design and evaluate Fact Check retrieval pipelines. Further, it provides a query interface for non-technical users to find relevant Fact Checks for the input query in the form of a key phrase, social media post, or an image.
• URL:
  https://­gitlab.­inria.­fr/­cedar/­factcheckbureau
• Publication:
  hal-04684068
• Contact:
  Ioana Manolescu

8.1.1 Graph integration of heterogeneous data sources for data journalism

Participants: Oana-Denisa Balalau, Nelly Barret, Theo Bouganim, Simon Ebel, Theo Galizzi, Ioana Manolescu, Madhulika Mohanty.

Work carried within the ANR AI Chair SourcesSay project has focused on developing a platform for integrating arbitrary heterogeneous data into a graph, then exploring and querying that graph using simple, intuitive query interfaces. The main technical challenges are: (i) how to interconnect structured and semi-structured data sources? We address this through information extraction (when an entity appears in two data sources or two places in the same graph, we only create one node, thus interlinking the two locations) and through similarity comparisons7.1.1; (ii) how to find all connections between nodes matching specific search criteria, or certain keywords? The question is particularly challenging in our context since ConnectionLens graphs can be pretty large, and query answers can traverse edges in both directions(iii) how to convert this graph into standard graph data models like property graphs, etc.

In this context, the following new contributions have been brought:

1. ConnectionStudio: a user-friendly data lake for data exploration. ConnectionStudio 16, 19, 27 is a user-friendly data lake, ingesting structured, semi-structured and un-structured documents (XML, JSON, CSV, RDF, PG, PDFs and Office files). It provides users different ways to explore and query the underlying data. For instance, it provides a set of lake-level statistics about the entities found in the data and Entity-Relationship diagrams showing the datasets structures. Further, it allows users to enumerate and browse a set of (interesting) paths connecting entities of interest in the data. It also provides a querying interface 7.1.4 to let users formulate their queries using few variables and inspect the data lake without having to write SQL queries. Finally, users can save, export and share results they produce in the lake, e.g., data journalists can share them in newsrooms.
2. Computing Generic Abstractions from Application Datasets The unprecedented creation, use and share of data have led to large, complex, more or less structured, and heterogeneous datasets in terms of model and schema. Therefore, users crucially need tools to consolidate and interact with sources gathered from various actors in order to compute practical results. Users of ConnectionLens7.1.1 who are not familiar with the graph formalism or who know very little about the structure and/or content of their datasets can quickly become lost. Therefore, in 21, we propose a novel generic method, named Abstra 7.1.2, to automatically generate data descriptions, in the form of diagrams like Entity-Relationship schemas, for structured and semi-structured data, such as JSON, XML, RDF and PG, without requiring to define new methods for each model.
3. Keyword Query Batch Evaluation on Graphs In this work, we consider the setting where many keyword queries are issued simultaneously. To solve them, one could launch an algorithm such as MoLESP 33 individually on each query. However, this approach may be inefficient, especially when some queries in the batch share common keywords. In such cases, computing the query results separately would repeat the exploration of paths starting from the common keywords. In this work, we develop algorithms for answering simultaneously several keyword queries without duplicating exploration work while utilising effective pruning techniques to stay practical with formally proven guarantees on the results.
4. Turning Heterogeneous Data Into Property Graphs Digital data proliferation has lead to the multiplication of data sources of various formats, organized in different ways, yet which may be used together. Following up on our ConnectionLens project which integrates any data sources into a simple graph, and Abstra which analyzes such graphs and casts them into Entity-Relationship diagrams, we have investigated the problem of automatically converting any (set of) datasets into a Property Graph. The benefits of such a conversion are: (a) a single format simplifies data processing, (b) Property Graphs tend to enable more efficient query processing by attaching attributes to nodes, which removes some of the joins needed by the processing of queries over RDF graphs; (c) there is now widespread research interest in Property Graphs, in particular leading to the development of a host of PG data storage and query processing systems. Towards this goal, we have described a preliminary approach for generating PG Schemas from Abstra data abstractions in 20. Separately, we have proposed a method for interactively visualizing any data source based on the ConnectionLens graph conversion followed by abstraction and finally PG conversion 28.

ConnectionLens is available online at: ConnectionLens Gitlab repository, while ConnectionStudio is available at ConnectionStudio Gitlab repository.

8.1.2 Decentralized semantic data sharing with access control

Participants: Maxime Buron, Hritika Kathuria, Ioana Manolescu, Georgios Siachamis.

Data exchange and sharing is a common need in virtually all modern applications. To achieve interoperability among different, heterogeneous databases, graph databases, and in particular knowledge graphs, are preferred due to their flexibility, which enables them to describe different structures of the underlying databases. Ontology-Based Data Access (OBDA) is the name commonly given to data integration systems based on knowledge graphs and ontologies; they have been successfully deployed in a variety of applications. Our team is a partner in DXP (Data Exchange Platform), a collaborative project between several Inria teams and Amadeus, technology provided for the travel industry. Within DXP, we are working to develop scalable, decentralized, and secure OBDA mechanisms for exchanging data across the different partners involved in a travel application: providers of services such as transport and lodging, travel operators, individual travelers, etc.

8.1.3 Finding Subgraphs with Maximum Total Density and Limited Overlap in Weighted Hypergraphs

Participants: Oana-Denisa Balalau.

Finding dense subgraphs in large (hyper)graphs is a key primitive in a variety of real-world application domains, encompassing social network analytics, event detection, biology, and finance. In most such applications, one typically aims at finding several (possibly overlapping) dense subgraphs, which might correspond to communities in social networks or interesting events. While a large amount of work is devoted to finding a single densest subgraph, perhaps surprisingly, the problem of finding several dense subgraphs in weighted hypergraphs with limited overlap has not been studied in a principled way, to the best of our knowledge. In our work 12, we define and study a natural generalization of the densest subgraph problem in weighted hypergraphs, where the main goal is to find at most k subgraphs with maximum total aggregate density, while satisfying an upper bound on the pairwise weighted Jaccard coefficient, i.e., the ratio of weights of intersection divided by weights of union on two nodes sets of the subgraphs. After showing that such a problem is NP-Hard, we devise an efficient algorithm that comes with provable guarantees in some cases of interest, as well as, an efficient practical heuristic. Our extensive evaluation on large real-world hypergraphs confirms the efficiency and effectiveness of our algorithms.

8.1.4 FactCheckBureau: Build Your Own Fact-Check Analysis Pipeline

Participants: Oana-Denisa Balalau, Garima Gaur, Oana Goga, Ioana Manolescu.

Fact-checkers are overwhelmed by the volume of claims they need to pay attention to fight misinformation. Even once debunked, a claim may still be spread by people unaware that it is false, or it may be recycled as a source of inspiration by malicious users. Hence, the importance of fact-check retrieval (FCR) as a research problem: given a claim and a database of previous checks, find the checks relevant to the claim. Existing solutions addressing this problem rely on the strategy of retrieve and re-rank relevant documents. Leveraging this pattern, we have built FactCheckBureau 17, an end-to-end solution for quickly designing, evaluating and deploying FCR pipelines. FactCheckBureau 7.1.8 serves the dual purpose – on one hand, researchers can interactively built and analyze their FCR pipelines and, on the other hand, non-technical users can use the FC query interface to fetch relevant fact checks from our corpus. Along with the tool, we also present fact check corpus we have built, which can be used in further research to test fact-check retrieval tools.

8.1.5 STaR: Space and Time-aware Statistic Query Answering

Participants: Oana-Denisa Balalau, Simon Ebel, Helena Galhardas, Theo Galizzi, Ioana Manolescu.

High-quality data is essential for informed public debate. Highquality statistical data sources provide valuable reference information for verifying claims. To assist journalists and fact-checkers, user queries about specific claims should be automatically answered using statistical tables. However, the large number and variety of these sources make this task challenging. STaR 18 is a novel method for Space and Time-aware STatistic Retrieval, based on a user natural language query. STaR is deployed within our system StatCheck, which we developed and shared with fact-checking journalists. STaR improves the quality of statistic fact retrieval by treating space and time separately from the other parts of the statistics dataset. Specifically, we use them as dimensions of the data (and the query), and focus the linguistic part of our dataset search on the rich, varied language present in the data. Our demonstration uses statistic datasets from France, Europe, and a few beyond, allowing users to query and explore along space and time dimensions.

8.1.6 Structured Discourse Representation for Factual Consistency Verification

Participants: Oana-Denisa Balalau, Ioana Manolescu, Kun Zhang.

Analysing differences between how events are represented in different texts, or verifying if large language models hallucinate, requires the ability to compare two texts. For this, a more structured representation of language can help by identifying distinct facts, and the discourse relationships connecting them, in each text. In this work, we first present an approach for extracting structured atomic facts and the relations between them, and then we present how to leverage this information to compare two texts. In addition, we show that the task of evaluating if the relation between facts has changed is challenging for SOTA LLMs. We believe that this work advances research on sentence similarity, by grounding it in linguistics and decomposing it in explainable modules. We demonstrate its potential in evaluating the factual consistency of two downstream tasks: text generation and text summarization.

8.1.7 Open information extraction for scientific text

Participants: Oana-Denisa Balalau, Garima Gaur, Ioana Manolescu, Prajna Upadhyay.

A conflict of interest occurs when a person or an organization is in a position to take actions or decisions in their official capacity to gain personal benefits—for instance, an oil company funds environmental researchers who are also on the company’s external advisory committee. In the interest of uncovering such incompatible situations, it is essential to have a knowledge base that captures relationships among different research studies, authors, and funding agencies. We work towards building such a knowledge base by leveraging the power of LLMs for the Open Information Extraction (OpenIE) from the scientific text. In particular, we focus on the acknowledgment sections or the metadata provided with the academic articles. We propose an end-to-end in-context learning-based framework to extract structured relationships from the unstructured text.

8.1.8 User intent modelling and information extraction in datalakes

Participants: Guillaume Lachaud, Fatemeh Nagesian.

Datalakes have the potential to help in many fields, with the ability to access rich and diverse information on a wide variety of topics. At the same time, the lack of structure inside a data lake make the joinability of the data a vital issue. Currently, data lakes remain vastly underexploited. In our work, we try to address this challenge in a two step process: (i) (ML for DB): we try to model user intent in an intermediate representation that can exploited by database systems; and (ii) (DB for ML): to answer a user query, we enrich the data with relevant information collected from the data lake. The user query can take the form of a machine learning model, such as a classification task.

8.2.1 Client-side and Server-side Tracking on Meta: Effectiveness and Accuracy

Growing concern over digital privacy has led to the widespread use of tracking restriction tools, such as ad blockers, Virtual Private Networks (VPN), and privacy-focused web browsers. All major browser vendors have also deprecated, or plan to deprecate, third-party cookies to reduce tracking. Despite these efforts, advertising companies continuously innovate to overcome these restrictions. Recently, advertising platforms, like Meta, have been promoting server-side tracking solutions to bypass traditional browser-based tracking restrictions. This paper explores how server-side tracking technologies can link website visitors with their user accounts on Meta products. The goal is to assess the effectiveness and accuracy of employing this technology, as well as the effect of tracking restrictions on online tracking. Our methodology involves a series of experiments where we integrate Meta's client-side tracker (the Meta Pixel) and server-side technology (the Conversions API) on different web pages. We then drive traffic to these pages and evaluate the success rate of linking website visitors to their profiles on Meta products. Our findings show that Meta's server-side technology can match between 34% and 51% of website visitors to user profiles on Meta products using basic information like the visitor's IP address, user agent, and location data. This is comparable to Pixel-based user matching in optimal conditions (i.e., in the absence of tracking restrictions), which links between 42% and 61% of user profiles. Nevertheless, we see a considerable difference in accuracy: while the Pixel-based tracking achieves 100% accuracy, less than 65% of the profiles matched by server-side tracking are accurate. The findings of this work 23 are extremely relevant in the current context of online tracking, where we recently saw Google allowing advertisers to use device fingerprinting to track and advertise to users. This paper has won the Andreas Pfitzmann Best Student Paper Award.

8.2.2 A Comparative Study of News Exposure and Consumption On and Off Facebook

Very large online platforms such as Meta, Google, and X use algorithms to feed users posts that match their interests. These algorithms have come under public scrutiny, as they can bias the information users consume in unwanted ways and consequently impact the formation of their political opinions and voting decisions. Many scholars have questioned whether social media promote misinformation and lead to echo chambers. To contribute to this debate, this study contrasts news exposure on Facebook (with an expected stronger algorithmic component) with news consumption off Facebook (with an expected weaker algorithmic component and stronger user behavior component) through the following questions: (1) Are users exposed to more/less misinformation on Facebook compared with their off-platform misinformation consumption? (2) Is news exposure on Facebook more/less diverse than off-platform news consumption? (3) To what extent do socio-demographic and psychological factors influence misinformation exposure and consumption both on and off Facebook? (4) Is there a relationship between socio-demographic and psychological factors and news diversity on and off Facebook? and (5) Does off-platform news consumption impact users’ exposure to misinformation on Facebook?

8.2.3 Analyzing Exposure and Consumption of News through Data Donations

Understanding how exposure to news on social media impacts public discourse and exacerbates political polarization is a significant endeavor in both computer and social sciences. Unfortunately, progress in this area is hampered by limited access to data due to the closed nature of social media platforms. Consequently, prior studies have been constrained to considering only fragments of users' news exposure and reactions. To overcome this obstacle, we present an innovative measurement 22 approach centered on donating personal data for scientific purposes, facilitated through a privacy-preserving tool that captures users' interactions with news on Facebook. This approach offers a nuanced perspective on users' news exposure and consumption, encompassing different types of news exposure: selective, incidental, algorithmic, and targeted, driven by the diverse underlying mechanisms governing news appearance on users' feeds. Our analysis of data from 472 participants based in the U.S. reveals several interesting findings. Furthermore, our study uncovers that users are open to engaging with news sources with opposite political ideology as long as these interactions are not visible to their immediate social circles. Overall, our study showcases the viability of data donation as a means to provide clarity to longstanding questions in this field, offering new perspectives on the intricate dynamics of social media news consumption and its effects.

8.2.4 NewsDB: An Automated Approach to Build an Extensive Database of Self-Proclaimed News Providers

The credibility of news obtained online has become a concern due to the ease with which individuals or groups can claim to be news publishers and share news-related content. Unfortunately, research on monitoring misleading information in the online news ecosystem is hindered because the community lacks a comprehensive and up-to-date list of social media pages and domains claiming to be news media. In this work, we propose an automated approach that uses Google's GNews API and Meta's CrowdTangle API (now replaced by Meta's Content Library API) to identify self-proclaimed news providers. Our method was able to discover 19k self-proclaimed news providers in the United States active in June 2022 and 23k active in October 2019. This is significantly more than the known 1,553 U.S.-based sources listed by Media Bias Fact Check and News Guard. We are currently compiling the largest database of self-proclaimed news providers worldwide. We have compiled 776 sources in the UK, 1629 in France, 1012 in Germany, 1188 in Italy, 2655 in Brazil, and 764 in India.

8.2.5 Privacy-oriented personalization of online services

The digital advertising industry has rapidly grown as a key revenue source not only for smaller content publishers and developers, such as mobile app creators, but also for tech giants like Google and Facebook. Over time, Advertisers have shifted towards behavioral advertising, where users are targeted based on their demographics, interests, and preferences. This approach relies heavily on extensive data collection to build user profiles, often conducted by third-party tracking scripts. But, these mechanisms have been repeatedly shown to come at the expense of user privacy with invasive tracking across websites. Therefore, web browsers are adopting privacy-preserving alternatives to existing advertising mechanisms, ensuring the advertising ecosystem can continue to operate effectively while preserving the user privacy. Third-Party Cookies (3PCs) have already been deprecated by browsers such as Safari, Brave, and Mozilla Firefox, while others, like Google Chrome, have announced plans to follow suit through a series of proposals under the “Privacy Sandbox Initiative”. At this stage, advertisers and publishers still have doubts regarding the effectiveness of these new technologies in targeting users compared to traditional 3PCs tracking and how they will affect their business. Conversely, it remains uncertain to what extent these technologies protect users privacy and respect their choices. Our work aims to analyze and evaluate the effectiveness, as well as assess the privacy risks associated with these new privacy-preserving advertising technologies.

8.3.1 Finding Conflicts of Opinion in Citizen Participation Platforms

Online citizen participation platforms are powerful democratic tools that allow large numbers of contributors to be involved in public decision-making. However, for large groups of contributors to collaborate, we need to provide tools for users and decision makers to navigate and understand high volumes of content. Towards this goal, we introduce in 15 an approach based on natural language processing to detect pairs of contradictory and equivalent proposals in online citizen participation contexts. We apply this approach on two major national citizen consultations: the République Numérique and Revenu Universel d'Activité consultations. We highlight the potential of our method in two use cases. First, our method is a high-quality tool for finding idea communities in online content. Second, we demonstrate that the method improves on the state-of-the-art for finding relevant complementary content for a user, by identifying new relevant views for 76% of the proposals test.

8.3.2 Navigating the Political Compass: Evaluating Multilingual LLMs Across Languages and Nationalities

Large Language Models (LLMs), particularly decoder models like GPT-4, Gemma, and LLaMA, have revolutionized applications ranging from creative writing and article summarization to political debate analysis. While these models excel in multilingual support, enabling accessibility to diverse user groups, concerns about inherent biases, including gender, stereotypical, and political biases, remain significant. This study specifically investigates political biases in multilingual LLMs, focusing on their tendencies toward particular political ideologies. We evaluate the political inclinations of these models using the Political Compass Test, examining the impact of model size, prompt language, and assigned nationality personas. Our findings reveal that the political biases of LLMs are influenced by these factors, highlighting the complex interplay between linguistic and cultural contexts in shaping model outputs.

8.3.3 Automatic detection of greenwashing

This work 25 focuses on the automatic detection of greenwashing using advanced NLP techniques, aiming to address the growing need for accountability in corporate communication. In collaboration with journalists, we explored real-world challenges and contextualized our efforts to ensure practical relevance. The project includes a comprehensive literature review, identifying key trends and gaps in existing approaches. A significant outcome is the development of a toolbox of NLP tools tailored for detecting greenwashing in corporate disclosures. Looking ahead, we aim to expand this research by developing a tool for analyzing corporate carbon transition plans, leveraging the capabilities of large language models (LLMs) to assess credibility and alignment with stated sustainability goals.

8.4.1 Efficient Version Space Algorithms for Human-in-the-loop Model Development

When active learning (AL) is applied to help the user develop a model on a large dataset through interactively presenting data instances for labeling, existing AL techniques often suffer from two main drawbacks: First, to reach high accuracy they may require hundreds of data instances to be labeled by the user, which is an onerous task for the user. Second, retrieving the next instance to label from a large dataset can be time- consuming, making it incompatible with the interactive nature of the human exploration process. To address these issues, we introduce a novel version-space-based active learner for kernel classifiers, which possesses strong theoretical guarantees on performance and efficient implementation in time and space. In addition, by leveraging additional insights obtained in the user labeling process, we can factorize the version space to perform active learning in a set of subspaces, which further reduces the user labeling effort. Evaluation results show that our algorithms 13 significantly outperform state-of-the-art version space strategies, as well as a recent factorization-aware algorithm, for model development over large data sets.

8.4.2 Efficient and robust active learning methods for interactive database exploration

There is an increasing gap between fast growth of data and the limited human ability to comprehend data. Consequently, therehas been a growing demand of data management tools that can bridge this gap and help the user retrieve high-value contentfrom data more effectively. In this work, we propose an interactive data exploration system as a new database service, usingan approach called “explore-by-example.” Our new system is designed to assist the user in performing highly effective dataexploration while reducing the human effort in the process. We cast the explore-by-example problem in a principled “activelearning” framework. However, traditional active learning suffers from two fundamental limitations: slow convergence andlack of robustness under label noise. To overcome the slow convergence and label noise problems, we bring the propertiesof important classes of database queries to bear on the design of new algorithms and optimizations for active learning-baseddatabase exploration. Evaluation results using real-world datasets and user interest patterns show that our new system, both inthe noise-free case and in the label noise case, significantly outperforms state-of-the-art active learning techniques and dataexploration systems in accuracy while achieving the desired efficiency for interactive data exploration.

8.4.3 A Spark Optimizer for Adaptive Fine-Grained Parameter Tuning

As Spark becomes a common big data analytics platform, its growing complexity makes automatic tuning of numerous parameters critical for performance. Our work on Spark parameter tuning is particularly motivated by two recent trends: Spark’s Adaptive Query Execution (AQE) based on runtime statistics, and the increasingly popular Spark cloud deployments that make cost-performance rea- soning crucial for the end user. This paper presents our design of a Spark optimizer that controls all tunable parameters of each query in the new AQE architecture to explore its performance benefits and, at the same time, casts the tuning problem in the theoretically sound multi-objective optimization (MOO) setting to better adapt to user cost-performance preferences. To this end, we propose a novel hybrid compile-time/runtime approach 26 to multi-granularity tuning of diverse, correlated Spark parameters, as well as a suite of modeling and optimization techniques to solve the tuning problem in the MOO setting while meeting the stringent time constraint of 1-2 seconds for cloud use. Evaluation results using TPC-H and TPC-DS benchmarks demonstrate the superior performance of our approach: (i) When prioritizing latency, it achieves 63% and 65% reduction for TPC-H and TPC-DS, respectively, under an average solving time of 0.7-0.8 sec, outperforming the most competitive MOO method that reduces only 18-25% latency with 2.6-15 sec solving time. (ii) When shifting preferences between latency and cost, our approach dominates the solutions of alternative methods, exhibiting superior adaptability to varying preferences.

8.4.4 Scalable Analytics on Multi-Streams Dynamic Graphs

Participants: Ioana Manolescu, Oana Balalau, Yanlei Diao, Ghufran Khan, Maxime Buron, Hritika Kathuria, Georgios Siachamis.

Several real-time applications rely on dynamic graphs to model and store data arriving from multiple streams. In addition to the high ingestion rate, the storage and query execution challenges are amplified in contexts where consistency should be considered when storing and querying the data. This work 30 addresses the challenges associated with multi-stream dynamic graph analytics. We propose a database design that can provide scalable storage and indexing, to support consistent read-only analytical queries (present and historical), in the presence of real-time dynamic graph updates that arrive continuously from multiple streams. This work has been accepted in VLDB 2025.

10.1.1 Associate Teams in the framework of an Inria International Lab or in the framework of an Inria International Program

10.2.1 Visits of international scientists

10.2.2 Visits to international teams

10.3.1 Horizon Europe

10.3.2 H2020 projects

Oana Goga is the Local PI for H2020 Trustaware (June 2021 - May 2024)

10.4.1 ANR

• Oana Goga is the local PI for LIX partner – ANR PRC 2022 – 2026 “FeedingBias: A multi-platform mixed-methods approach to news exposure on social media” (our part: 128,000 €)
• Oana Goga is the local PI for LIX partner – ANR PRCE 2021 – 2025 “PROPEOS: Privacy-oriented Personalization of Online Services” (our part: 202,720 €)
• Ioana Manolescu is the local PI for Inria Saclay in CQFD (2019-2024), an ANR project coordinated by F. Ulliana (U. Montpellier). Its research aims at investigating efficient data management methods for ontology-based access to heterogeneous databases (polystores).
• Ioana Manolescu is the PI of SourcesSay (2020-2024), an AI Chair funded by Agence Nationale de la Recherche and Direction Générale de l’Armement. The project goal is to interconnect data sources of any nature within digital arenas. In an arena, a dataset is stored, analyzed, enriched and connected, graph mining, machine learning, and visualization techniques, to build powerful data analysis tools.

10.4.2 Others

• Associated Inria team MediumAI was accepted for a period of 3 years (2024-2026), PI: Oana Balalau and Davide Ceolin (CWI, Amsterdam).
• A 2-years post-doc funding was obtained from Hi!Paris for Garima Gaur with an application by Ioana Manolescu , Oana Balalau , and Oana Goga .
• A 2-years post-doc funding was obtained from Inria for Chadi Helwe with an application by Oana Balalau and Davide Ceolin (CWI, Amsterdam).

11.1.1 Scientific events: organisation

11.1.2 Scientific events: selection

11.1.3 Journal

11.1.4 Invited talks

The team members have given invited talks at various venues to national and international audience.

Oana Balalau has delivered the following talks:

• Keynote at InfoxSurSeine 2024, Paris
• Presentation and round table at I-EXPO 2024, Paris
• Presentation at the European Cyber Week, in the Cyber cognitive threats seminar, Rennes
• Presentation at LIX seminar, Palaiseau
• Presentation at the AI & Social Sciences Seminar, CREST, Palaiseau

Oana Goga has delivered the following national (N) and international (I) talks:

• (N) Plenary talk at Journée Inria - Ministére des Armées, June 2024
• (N) Plenary talk at Congrès Annuel de la Société Informatique de France, June 2024
• (I) Plenary talk at CNIL Privacy Research Day, June 2024
• (N) Plenary talk at Academie des Sciences, May 2024
• (N) Panel at CPDP The Role of Research and Researchers in AI Governance, May 2024
• (I) Plenary talk Meeting of the OECD Working Party on Digital Economics, Measurement, and Analysis, April 2024
• (I) Plenary talk at Forum International de la Cybersécurité – FIC 2024, March 2024
• (I) Panel at OECD report release Facts not Fakes: Tackling Disinformation, Strengthening Information Integrity, March 2024
• (I) Plenary talk at DSA and Platform Regulation Conference, Feb 2024
• (N) Plenary talk at Musée des Arts et Métiers – L'aventure des inventions, Feb 2024, general audience

Ioana Manolescu has delivered the following talks:

• Invited keynote at the ISWC 2024 conference: “Retrieve (and Leverage)the Inner Graph Behind the Data”
• Invited keynote at the GRADES 2024 workshop in conjunction with SIGMOD 2024: “The Real World is Graph-Structured: Retrieving Meaning from Heterogeneous Data” 14
• Keynote at the Junior Data Science and Engineering Conference at U. Paris Saclay: “Data and AI techniques for Fact-checking and Investigative journalism”
• Invited talk at a Colloquium on Conflict of Interests at Sciences Po Paris, January 2024: “Intégration de données hétérogènes et détection de liens/conflits d'intérêt: Système ConnectionLens”
• Invited talk at the EU Joint Research Center (JRC) Workshop on Disinformation, September 2024: “Harnessing data and AI to catch and unmask fakes”

Yanlei Diao has delivered an invited keynote at AWS Machine Learning Workshop on “Transformers and Foundation Models for Big Data Analytics Systems”, September 2024.

11.1.5 Leadership within the scientific community

Ioana Manolescu has been the president of the informal French Data Management Association (BDA).

11.1.6 Scientific expertise

Oana Goga has provided her scientific expertise under the following roles:

• Member of the Science Advisory Committee for the NSF-funded Mid-scale RI-1 project (a 15 million US dollars project): Observatory for online human and platform behavior (2024).
• Member of the OECD Task Force on Global Challenge to Build Trust in the Age of Generative AI (2024).
• External Expert with contract for the European Commission (DG Connect & ECAT), to advise on the Delegated Act addressing the implementation of the DSA Article 40 that gives vetted researchers access to data from very large online platforms and very large search engines (2024).

11.1.7 Research administration

Ioana Manolescu has contributed to the following:

• She represents Inria in the Comité Operationnel of Hi!Paris, an AI Pole of Excellency comprising IP Paris and HEC.
• She is an elected member of IP Paris' Comité Académique and serves on its Scientific Committee.
• She is the head of the Data Analytics and Machine Learning axis of research within LIX.

11.2.1 Teaching

Oana Balalau is a part-time (33%) assistant professor at Ecole Polytechnique, where she teaches in two courses:

• INF473G “Mining, learning and reasoning on Web Graphs”, L3, Ecole Polytechnique
• INF583 “Systems for Big Data”, M1, Ecole Polytechnique

Yanlei Diao holds a part-time (50%) position at Ecole Polytechnique. She teaches INF583 “Systems for Big Data ”, M1, Ecole Polytechnique

Garima Gaur taught 6h TP for the course ECE5DA04TP “Big Graph Databases”, M1/M2, Télécom

Ioana Manolescu is a part-time professor (50%) at Ecole Polytechnique. She taught:

• Courses, labs and TDs in CSC51053EP “Database Management Systems”, M1, Ecole Polytechnique
• She is in charge of the M1 Internship program in Artificial Intelligence and Data Science (CSC51092EP).
• She is also in charge of the Artificial Intelligence and Data Science program (M1) at Ecole Polytechnique
• Further, Ioana Manolescu has taught 9h of lectures for the new course ECE5DA04TP “Big Graph Databases”, M1/M2, Télécom.

Madhulika Mohanty has taught:

• 3h of lectures and 3h of TP for the course ECE5DA04TP “Big Graph Databases”, M1/M2, Télécom.
• 28 h of lectures and 28h of labs for INF540 “Databases”, M1, Ecole Polytechnique.
• 18h of TP for CSC51053EP “Database Management Systems”, M1, Ecole Polytechnique

11.2.2 Supervision

11.2.3 Juries

The team members have served on various juries.

Oana Balalau was a:

• member of the recruitment committee for part-time teaching position for Ecole Polytechnique
• member of the recruitment comittee for assistant professor at Ecole Polytechnique
• part of the PhD defense committee of Hadi Abdine, Ecole Polytechnique

Oana Goga was the part of the following committees:

• Award comitee: CNIL-Inria Privacy Protection Award 2024
• Selection comitee: tenure committee for Mobin Javed at LUMS (2024), Assistant Professor for Telecom ParisTech (2024)
• Ph.D. Committee jury member: Paul Bouchaud (2024), Evan Dufraisse (September 2024), Alireza Mohammadinodooshan (June 2024), Naif Mehanna (May 2024), Moitree Basu (Jan 2024)

Ioana Manolescu reported on the Ph.D. of Alexandra Rogova (co-supervised by Leonid Libkin and Amélie Gheerbrant), U. Paris 6, defended in November 2024.

11.3.1 Internal or external Inria responsibilities

• Oana Balalau is a member of Inria Saclay's Scientific Commission.
• Oana Goga is a member of Conseil Scientifique of Regalia (2024).
• Ioana Manolescu is an elected member of Inria's Comité d'Evaluation.

11.3.2 Scientific Interventions

• Oana Goga appeared in an interview for CNRS Bronze Medal Award: “Oana Goga récompensée pour l’étude des risques liés aux plateformes en ligne” in March 2024.
• Oana Goga delivered a seminar at Musée des Arts et Métiers – L’aventure des inventions in February 2024.
• Oana Balalau participated in “L'intelligence artificielle au service du bien public”, for young students (seconde) during their Inria internship, 2024.
• The Franco-British Council in partnership with Inria and the Franco-British Data Society hosted a one-day seminar on “Artificial Intelligence and Information Warfare” in July 2024 at University College London. Simon Ebel and Theo Galizzi presented StatCheck and Garima Gaur presented FactCheckBureau. Garima Gaur also participated in a panel discussion with Benjamin Guedj (Director of Research at Inria, Professor of Machine Learning and Foundational AI at University College London, and Turing Fellow). The full report is available here.
• Nelly Barret delivered a talk on “User-friendly data lake exploration” at LIB (Université de Dijon) in Jan. 2024.
• Nelly Barret talked about “Des données au journalisme” at Lycée international de Palaiseau Paris Saclay (LIPPS), in Jan. 2024.
• Salim Chouaki delivered a talk on “Analyzing Risks with Information Exposure on Social Media” at National Taipei University (Taiwan) in November 2024.
• Madhulika Mohanty has delivered talks on “Effective Exploration of Graph-Structured Data” at LIFO, Orléans and at the BOREAL team of Inria at Montpellier in January 2024.

International journals

• 12 articleO.Oana Balalau, F.Francesco Bonchi, T.-H. H.T-H. Hubert Chan, F.Francesco Gullo, M.Mauro Sozio and H.Hao Xie. Finding Subgraphs with Maximum Total Density and Limited Overlap in Weighted Hypergraphs.ACM Transactions on Knowledge Discovery from Data (TKDD)184February 2024, 1-21HALDOIback to text
• 13 articleL.Luciano Di Palma, Y.Yanlei Diao and A.Anna Liu. Efficient Version Space Algorithms for Human-in-the-loop Model Development.ACM Transactions on Knowledge Discovery from Data (TKDD)183January 2024, 1-49HALDOIback to text

Invited conferences

• 14 inproceedingsI.Ioana Manolescu. The Real World is Graph-Structured: Retrieving Meaning from Heterogeneous Data (invited keynote).GRADES-NDA 2024 - 7th Joint Workshop on Graph Data Management Experiences & Systems (GRADES) and Network Data Analytics (NDA)Santiago (CL), ChileJune 2024HALback to text

International peer-reviewed conferences

• 15 inproceedingsW.William Aboucaya, O.Oana Balalau, R.Rafael Angarita and V.Valérie Issarny. Finding Conflicts of Opinion in Citizen Participation Platforms.2024 IEEE International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT)Bangkok, ThailandDecember 2024, 8HALback to text
• 16 inproceedingsO.Oana Balalau, N.Nelly Barret, S.Simon Ebel, T.Théo Galizzi, I.Ioana Manolescu and M.Madhulika Mohanty. Graph lenses over any data: the ConnectionLens experience.ICDE 2024 - 40th IEEE International Conference on Data EngineeringUtrecht, NetherlandsMay 2024HALback to text
• 17 inproceedingsO.Oana Balalau, P.Pablo Bertaud-Velten, Y.Younes El-Fraihi, G.Garima Gaur, O.Oana Goga, S.Samuel Guimaraes, I.Ioana Manolescu and B.Brahim Saadi. FactCheckBureau: Build Your Own Fact-Check Analysis Pipeline.CIKM 2024 - 33rd ACM International Conference on Information and Knowledge ManagementBoise Idaho, United StatesOctober 2024HALDOIback to text
• 18 inproceedingsO.Oana Balalau, S.Simon Ebel, H.Helena Galhardas, T.Théo Galizzi and I.Ioana Manolescu. STaR: Space and Time-aware Statistic Query Answering.ACM Conference on Information and Knowledge Management (CIKM)CIKM 2024 - 33rd ACM International Conference on Information and Knowledge ManagementBoise, Idaho, United StatesOctober 2024HALDOIback to text
• 19 inproceedingsN.Nelly Barret, S.Simon Ebel, T.Théo Galizzi, I.Ioana Manolescu and M.Madhulika Mohanty. Exploration utilisateur de lacs de données très hétérogènes.EGC 2024 - 24ème conférence francophone sur l'Extraction et la Gestion des ConnaissancesDijon, FranceJanuary 2024HALback to text
• 20 inproceedingsN.Nelly Barret, T.Tudor Enache, I.Ioana Manolescu and M.Madhulika Mohanty. Finding the PG schema of any (semi)structured dataset: a tale of graphs and abstraction.SEAGraph Workshop 2024 - 3rd Workshop on Search, Exploration, and Analysis in Heterogeneous Datastores - 40th IEEE International Conference on Data Engineering (ICDE 2024)Utrecht, Netherlands2024HALback to text
• 21 inproceedingsN.Nelly Barret, I.Ioana Manolescu and P.Prajna Upadhyay. Computing Generic Abstractions from Application Datasets.OpenProceedingsEDBT 2024 - 27th International Conference on Extending Database Technology27Paestum, ItalyMarch 2024, 94-107HALback to text
• 22 inproceedingsS.Salim Chouaki, A.Abhijnan Chakraborty, O.Oana Goga and S.Savvas Zannettou. What News Do People Get on Social Media? Analyzing Exposure and Consumption of News through Data Donations.WWW 2024 - The ACM International World Wide Web ConferenceSingapore, SingaporeACMMay 2024, 2371-2382HALDOIback to text
• 23 inproceedingsA.Asmaa El fraihi, N.Nardjes Amieur, W.Walter Rudametkin and O.Oana Goga. Client-side and Server-side Tracking on Meta: Effectiveness and Accuracy.Proceedings on Privacy Enhancing TechnologiesPETS 2024 - 24th Privacy Enhancing Technologies Symposium20243Bristol, United KingdomJuly 2024, 431-445HALDOIback to text
• 24 inproceedingsA.Antoine Gauquier. Towards Efficient Construction of a Traceable, Multimodal, and Heterogeneous Data Warehouse.CEUR Workshop ProceedingsVLDB 2024 PhD Workshop - The 50th International Conference on Very Large Data BasesGuangzhou, ChinaAugust 2024HAL
• 25 inproceedingsC.Chadi Helwe, T.Tom Calamai, P.-H.Pierre-Henri Paris, C.Chloé Clavel and F. M.Fabian M Suchanek. MAFALDA: A Benchmark and Comprehensive Study of Fallacy Detection and Classification.Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language TechnologiesNAACL 2024 - North American Chapter of the Association for Computational Linguistics1: Long PapersMexico City, MexicoAssociation for Computational LinguisticsJune 2024, 4810-4845HALDOIback to text
• 26 inproceedingsC.Chenghao Lyu, Q.Qi Fan, P.Philippe Guyard and Y.Yanlei Diao. A Spark Optimizer for Adaptive, Fine-Grained Parameter Tuning.VLDB 2024 - 50th International Conference on Very Large DatabasesGuangzhou, ChinaAugust 2024HALDOIback to text

Conferences without proceedings

• 27 inproceedingsN.Nelly Barret, N.Nikola Dobričić, S.Simon Ebel, T.Théo Galizzi, I.Ioana Manolescu and M.Madhulika Mohanty. ConnectionStudio: User-friendly Exploration of Highly Heterogeneous Data Lakes.BDA « Gestion de Données – Principes, Technologies et Applications »Orleans (France), FranceOctober 2024HALback to text
• 28 inproceedingsT.Théo Bouganim, I.Ioana Manolescu and E.Emmanuel Pietriga. A Unified Visual Exploration Framework for (Semi-)structured Data.BigVis 2024 - 7th International Workshop on Big Data Visual Exploration and Analytics (VLDB 2024)Guangzhou, ChinaAugust 2024HALback to text

Doctoral dissertations and habilitation theses

• 29 thesisN.Nelly Barret. User-oriented exploration of semi-structured datasets.Institut Polytechnique de ParisMarch 2024HALback to text

Reports & preprints

• 30 reportA. C.Angelos Christos Anadiotis, M.Muhammad Ghufran Khan and I.Ioana Manolescu. Dynamic Graph Databases with Out-of-order Updates (extended version).Institut Polytechnique de Paris; INRIAOctober 2024HALback to text

Scientific popularization

• 31 miscB.Bram Vanderborght, M.Miguel Colom, R.Rūta Binkytė, O.Oana Balalau, O.Oana Goga, H.Hervé Debar, M.Marceau Coupechoux and J.Juan Herrera. E.Eva Poptcheva, eds. Alvolution - al and digital technologies in the European Union: Conclusions.April 2024HAL