Jean-Gabriel Ganascia

Camilo Sarmiento, Gauvain Bourgne, Katsumi Inoue et al.

Although moral responsibility is not circumscribed by causality, they are both closely intermixed. Furthermore, rationally understanding the evolution of the physical world is inherently linked with the idea of causality. Thus, the decision-making applications based on automated planning inevitably have to deal with causality, especially if they consider imputability aspects or integrate references to ethical norms. The many debates around causation in the last decades have shown how complex this notion is and thus, how difficult is its integration with planning. As a result, much of the work in computational ethics relegates causality to the background, despite the considerations stated above. This paper's contribution is to provide a complete and sound translation into logic programming from an actual causation definition suitable for action languages, this definition is a formalisation of Wright's NESS test. The obtained logic program allows to deal with complex causal relations. In addition to enabling agents to reason about causality, this contribution specifically enables the computational ethics domain to handle situations that were previously out of reach. In a context where ethical considerations in decision-making are increasingly important, advances in computational ethics can greatly benefit the entire AI community.

Evangelia Zve, Gauvain Bourgne, Benjamin Icard et al.

Outliers in dynamic topic modeling are typically treated as noise, yet we show that some can serve as early signals of emerging topics. We introduce a temporal taxonomy of news-document trajectories that defines how documents relate to topic formation over time. It distinguishes anticipatory outliers, which precede the topics they later join, from documents that either reinforce existing topics or remain isolated. By capturing these trajectories, the taxonomy links weak-signal detection with temporal topic modeling and clarifies how individual articles anticipate, initiate, or drift within evolving clusters. We implement it in a cumulative clustering setting using document embeddings from eleven state-of-the-art language models and evaluate it retrospectively on HydroNewsFr, a French news corpus on the hydrogen economy. Inter-model agreement reveals a small, high-consensus subset of anticipatory outliers, increasing confidence in these labels. Qualitative case studies further illustrate these trajectories through concrete topic developments.

Benjamin Icard, Lila Sainero, Alice Breton et al.

Large language models (LLMs) can convincingly imitate human writing styles, yet it remains unclear how much stylistic information is encoded in embeddings from any language model and retained after LLM rewriting. We investigate these questions in French, using a controlled literary dataset to quantify the effect of stylistic variation via changes in embedding dispersion. We observe that embeddings reliably capture authorial stylistic features and that these signals persist after rewriting, while also exhibiting LLM-specific patterns. These analytical results offer promising directions for authorship imitation detection in the era of language models.

Evangelia Zve, Benjamin Icard, Alice Breton et al.

This paper examines how outliers, often dismissed as noise in topic modeling, can act as weak signals of emerging topics in dynamic news corpora. Using vector embeddings from state-of-the-art language models and a cumulative clustering approach, we track their evolution over time in French and English news datasets focused on corporate social responsibility and climate change. The results reveal a consistent pattern: outliers tend to evolve into coherent topics over time across both models and languages.

Benjamin Icard, Evangelia Zve, Lila Sainero et al.

This paper analyzes how writing style affects the dispersion of embedding vectors across multiple, state-of-the-art language models. While early transformer models primarily aligned with topic modeling, this study examines the role of writing style in shaping embedding spaces. Using a literary corpus that alternates between topics and styles, we compare the sensitivity of language models across French and English. By analyzing the particular impact of style on embedding dispersion, we aim to better understand how language models process stylistic information, contributing to their overall interpretability.

Hussam Hamdan, Jean-Gabriel Ganascia

Many Entity Linking systems use collective graph-based methods to disambiguate the entity mentions within a document. Most of them have focused on graph construction and initial weighting of the candidate entities, less attention has been devoted to compare the graph ranking algorithms. In this work, we focus on the graph-based ranking algorithms, therefore we propose to apply five centrality measures: Degree, HITS, PageRank, Betweenness and Closeness. A disambiguation graph of candidate entities is constructed for each document using the popularity method, then centrality measures are applied to choose the most relevant candidate to boost the results of entity popularity method. We investigate the effectiveness of each centrality measure on the performance across different domains and datasets. Our experiments show that a simple and fast centrality measure such as Degree centrality can outperform other more time-consuming measures.

Suzanne Mpouli, Jean-Gabriel Ganascia

Similes play an important role in literary texts not only as rhetorical devices and as figures of speech but also because of their evocative power, their aptness for description and the relative ease with which they can be combined with other figures of speech (Israel et al. 2004). Detecting all types of simile constructions in a particular text therefore seems crucial when analysing the style of an author. Few research studies however have been dedicated to the study of less prominent simile markers in fictional prose and their relevance for stylistic studies. The present paper studies the frequency of adjective and verb simile markers in a corpus of British and French novels in order to determine which ones are really informative and worth including in a stylistic analysis. Furthermore, are those adjectives and verb simile markers used differently in both languages?

Suzanne Mpouli, Jean-Gabriel Ganascia

The present study is focused on the automatic identification and description of frozen similes in British and French novels written between the 19 th century and the beginning of the 20 th century. Two main patterns of frozen similes were considered: adjectival ground + simile marker + nominal vehicle (e.g. happy as a lark) and eventuality + simile marker + nominal vehicle (e.g. sleep like a top). All potential similes and their components were first extracted using a rule-based algorithm. Then, frozen similes were identified based on reference lists of existing similes and semantic distance between the tenor and the vehicle. The results obtained tend to confirm the fact that frozen similes are not used haphazardly in literary texts. In addition, contrary to how they are often presented, frozen similes often go beyond the ground or the eventuality and the vehicle to also include the tenor.

Jean-Gabriel Ganascia, Pierre Glaudes, Andrea Del Lungo

For more than forty years now, modern theories of literature (Compagnon, 1979) insist on the role of paraphrases, rewritings, citations, reciprocal borrowings and mutual contributions of any kinds. The notions of intertextuality, transtextuality, hypertextuality/hypotextuality, were introduced in the seventies and eighties to approach these phenomena. The careful analysis of these references is of particular interest in evaluating the distance that the creator voluntarily introduces with his/her masters. Phoebus is collaborative project that makes computer scientists from the University Pierre and Marie Curie (LIP6-UPMC) collaborate with the literary teams of Paris-Sorbonne University with the aim to develop efficient tools for literary studies that take advantage of modern computer science techniques. In this context, we have developed a piece of software that automatically detects and explores networks of textual reuses in classical literature. This paper describes the principles on which is based this program, the significant results that have already been obtained and the perspectives for the near future.