Papers by Olivier Ferret

16 papers
“Women do not have heart attacks!” Gender Biases in Automatically Generated Clinical Cases in French (2025.findings-naacl)

Copied to clipboard

Challenge: Healthcare professionals are increasingly including Language Models (LMs) in clinical practice.
Approach: They propose to use LMs to generate clinical cases in french and an automatic linguistic gender detection tool to measure gender biases.
Outcome: The proposed model over-generates cases describing male patients, creating synthetic corpora that are not consistent with documented prevalence for these disorders.
Extrinsic Evaluation of French Dependency Parsers on a Specialized Corpus: Comparison of Distributional Thesauri (2020.lrec-1)

Copied to clipboard

Challenge: Using a frequency-based method, we can identify subsets of the same word contexts without any reference data.
Approach: They compare 11 different French dependency parsers on a specialized corpus to generate distributional thesauri using a frequency-based method.
Outcome: The proposed method can identify relevant subsets without reference data and the similarity is confirmed on a restricted distributional benchmark.
Embedding Strategies for Specialized Domains: Application to Clinical Entity Recognition (P19-2)

Copied to clipboard

Challenge: Off-the-shelf word embeddings tend to perform poorly on texts from specialized domains such as clinical reports.
Approach: They combine off-the-shelf contextual embeddings with static word2vec embedders trained on a small in-domain corpus built from task data to reach and sometimes outperform representations learned from a large corpus in the medical domain.
Outcome: The proposed embedding strategies outperform representations learned from a large corpus in the medical domain.
Few-Shot Event Argument Extraction Based on a Meta-Learning Approach (2024.naacl-srw)

Copied to clipboard

Challenge: Recent studies on few-shot event extraction focus on event trigger detection and argument extraction in meta-learning contexts.
Approach: They propose to use prototypical networks to perform few-shot event argument extraction . they propose to inject syntactic knowledge into the model to enhance relation embeddings .
Outcome: The proposed approach achieves strong performance on ACE 2005 in several few-shot configurations.
Factually Consistent Summarization via Reinforcement Learning with Textual Entailment Feedback (2023.acl-long)

Copied to clipboard

Challenge: Recent advances in abstractive summarization systems produce factually inconsistent text . this is emphasized in tasks like summarizing, which often produce inconsistent text with no input article .
Approach: They use reinforcement learning to optimize for factual consistency and explore trade-offs . they use textual-entailment rewards to optimize the accuracy of the generated summaries .
Outcome: The proposed method improves faithfulness, salience and conciseness of the generated summaries.
Can Synthetic Text Help Clinical Named Entity Recognition? A Study of Electronic Health Records in French (2023.eacl-main)

Copied to clipboard

Challenge: In sensitive domains, the sharing of corpora is restricted due to confidentiality, copyrights or trade secrets.
Approach: They use auto-regressive neural models to generate a clinical case corpus annotated with clinical entities and evaluate it for a named entity recognition task.
Outcome: The proposed model can produce clinical case corpus annotated with clinical entities while maintaining confidentiality.
Building a Multimodal Entity Linking Dataset From Tweets (2020.lrec-1)

Copied to clipboard

Challenge: Entity linking is a task that aims at associating an entity mention with a unique entity in a knowledge base.
Approach: They propose a method to quasi-automatically build annotated datasets to evaluate methods on the Entity Linking task.
Outcome: The proposed method builds annotated datasets of tweets with ambiguous mentions and a Twitter KB defining the entities.
Building Static Embeddings from Contextual Ones: Is It Useful for Building Distributional Thesauri? (2022.lrec-1)

Copied to clipboard

Challenge: contextual language models are dominant in the field of Natural Language Processing, but they are not suitable for all uses.
Approach: They propose a method for building word or type-level embeddings from contextual models . they evaluate a large set of English nouns from the perspective of extracting semantic similarity relations .
Outcome: The proposed method can be used to build word or type embeddings from contextual models . it can be exploited for a wide set of English nouns, showing it can improve distributional thesauri .
Re-train or Train from Scratch? Comparing Pre-training Strategies of BERT in the Medical Domain (2022.lrec-1)

Copied to clipboard

Challenge: Recent years have witnessed the widespread use of transfer learning techniques in Natural Language Processing (NLP)
Approach: They train BERT models from scratch using many configurations involving general and medical corpora.
Outcome: The initial corpus only has a weak influence when these are further pre-trained on a medical corpus.
Conditional Language Policy: A General Framework For Steerable Multi-Objective Finetuning (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing approaches for multi-objective Reinforcement Learning (RL) are difficult due to plurality of preferences and applications.
Approach: They propose a framework for finetuning language models on multiple objectives using conditional language policy.
Outcome: The proposed framework outperforms and Pareto-dominates existing approaches for multi-objective Reinforcement Learning (RL) it does not require training or maintaining multiple models to achieve different trade-offs between the objectives.
Can We Guide a Multi-Hop Reasoning Language Model to Incrementally Learn at Each Single-Hop? (2022.coling-1)

Copied to clipboard

Challenge: Recent developments have shown that pre-trained language models are effective soft reasoners over language.
Approach: They propose to model multi-hop reasoning process as a sequence of explicit single-hop steps.
Outcome: The proposed model improves on multiple-choice question answering and reading comprehension with 68.4% and 16.0% w.r.t. classic PLMs.
CLISTER : A Corpus for Semantic Textual Similarity in French Clinical Narratives (2022.lrec-1)

Copied to clipboard

Challenge: Modern Natural Language Processing relies on the availability of annotated corpora for training and evaluation.
Approach: They propose to annotate sentences in French using a definition of similarity guided by clinical facts and use it to evaluate the corpus.
Outcome: The proposed model can capture similarity with state-of-the-art performance on the DEFT STS shared task evaluation data set.
CharacterBERT: Reconciling ELMo and BERT for Word-Level Open-Vocabulary Representations From Characters (2020.coling-main)

Copied to clipboard

Challenge: Pre-trained language representations from Transformers have become the most popular choice for building NLP systems.
Approach: They propose a new variant of BERT that drops the wordpiece tokenization system altogether . they propose 'characterBERT' module to represent entire words by consulting their characters .
Outcome: The proposed model improves performance on a variety of medical domain tasks while producing robust, word-level, and open-vocabulary representations.
Using pseudo-senses for improving the extraction of synonyms from word embeddings (P18-2)

Copied to clipboard

Challenge: Existing methods for specializing word embeddings rely on external knowledge . we propose a new method that focuses on semantic similarity and synonym extraction without external knowledge.
Approach: They propose a method that exploits pseudo-sense for building several representations for each word and uses these representations to make initial embeddings more generic.
Outcome: The proposed method improves word embeddings without external knowledge and focuses on semantic similarity and synonym extraction.
Language Models and Semantic Relations: A Dual Relationship (2024.lrec-main)

Copied to clipboard

Challenge: Existing studies on language models for the extraction of semantic relations have focused on injecting semantic knowledge into these models to enhance them.
Approach: They propose to extract lexical semantic relations from a BERT model and inject them into it using unsupervised methods based on semantic similarity at word and sentence levels.
Outcome: The proposed method allows to enrich a BERT model without using any external semantic resource.
Multi-Level Information Retrieval Augmented Generation for Knowledge-based Visual Question Answering (2024.emnlp-main)

Copied to clipboard

Challenge: Knowledge-Aware Visual Question Answering about Entity tasks require two separate steps to generate accurate answers.
Approach: They propose a multi-level information RAG approach that enhances answer generation through entity retrieval and query expansion.
Outcome: The proposed approach improves answer generation through entity retrieval and query expansion.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations