Papers by Christophe Cerisara
Improving Quotation Attribution with Fictional Character Embeddings (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent methods to attribute quotes to human logic lack character representations, which often leads to errors in more challenging examples of attribution: anaphoric and implicit quotes. |
| Approach: | They propose to augment a popular quotation attribution system, BookNLP, with character embeddings that encode global stylistic information of characters derived from an off-the-shelf stylometric model, Universal Authorship Representation (UAR). |
| Outcome: | The proposed system improves anaphoric and implicit quotes, reaching state-of-the-art. |
Efficient One-shot Compression via Low-Rank Local Feature Distillation (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing structured pruning approaches for large language models require calibration data and costly continued pretraining on billions of tokens to recover lost performance. |
| Approach: | They propose a method that locally distills activations with low-rank weights . they compress Mixtral-8x7B on a single GPU and Phi-2 3B by 40% . |
| Outcome: | The proposed method compresses Mixtral-8x7B on a single A100 GPU, removing 10 billion parameters while retaining over 95% of its original performance. |
Cross-lingual Matryoshka Representation Learning across Speech and Text (2026.findings-acl)
Copied to clipboard
| Challenge: | Speakers of under-represented languages face language barriers and modality barriers . we train a bilingual speech-text embedding model for French-Wolof . |
| Approach: | They train a bilingual speech-text Matryoshka embedding model that enables efficient retrieval of French text from Wolof speech queries. |
| Outcome: | The proposed model can retrieve French text from Wolof speech queries without expensive ASR-translation pipelines. |
Computational Narrative Understanding for Expressive Text-to-Speech (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in text-to-speech systems have been driven by large, multi-domain speech corpora. |
| Approach: | They propose a large-scale 5.3K hours of expressive speech drawn from character quotations . they fine-tune a flow-matching model and train from scratch . |
| Outcome: | The proposed model improves expressivity and intelligibility while training from scratch improves expressiveness of an autoregressive model. |
Multi-task dialog act and sentiment recognition on Mastodon (C18-1)
Copied to clipboard
| Challenge: | Social media are a gold mine for researchers in many domains and especially in natural language processing . license restrictions make it difficult to strictly reproduce research results on Twitter data . |
| Approach: | They propose to annotate a Twitter-like corpus from a decentralized social network with permissive licenses that are compatible with reproducible experiments. |
| Outcome: | The proposed method shows that transfer learning can be efficiently achieved between tasks. |
Unsupervised multiple-choice question generation for out-of-domain Q&A fine-tuning (2022.acl-short)
Copied to clipboard
| Challenge: | Pre-trained models have shown very good performances on a number of question answering benchmarks especially when fine-tuned on multiple question answering datasets at once. |
| Approach: | They propose an approach to fine-tune a question-answering dataset using a rule-based algorithm that generates questions and answers from unannotated sentences. |
| Outcome: | The proposed model can generate questions and answers from unannotated sentences on a multiple-choice physics, biology and chemistry benchmark. |
Evaluating LLMs for Quotation Attribution in Literary Texts: A Case Study of LLaMa3 (2025.naacl-short)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have shown promising results in literary tasks . however, quotation attribution remains a challenging task and methods that generalize across writing styles are lacking analysis regarding book memorization and annotation contamination. |
| Approach: | They evaluate the ability of Llama-3 to attribute utterances of direct-speech to their speaker in novels by assessing the impact of book memorization and annotation contamination. |
| Outcome: | The proposed model outperforms existing models on a corpus of 28 novels and shows that book memorization and annotation contamination do not explain the performance gain. |