software

  • PACT: pruning and clustering-based visual token reduction for faster vision-language models (CVPR 2025), with LLaVA-OneVision and Qwen2-VL backends
  • MBartCopyGenerator: mBART-based model for keyword indexing of scientific articles, with a BARThez version adapted to French (NLDB 2022)
  • SKG builder: tool to generate a Scientific Knowledge Graph from academic papers
  • TextAnnot: semantic annotation and IR (PostDoc project)
  • Kleis: CRF-based keyphrase extractor for scientific literature, also on PyPI
  • YaSemIR: Yet Another Semantic Information Retrieval system
  • SemEval / *SEM STS task participations: code used in the STS tasks 2012 and 2013