Events　represent　fundamental　constituents　of　the　world,　and　investigating　expressions　of　opinion　centered　around　events　can　enhance　our　comprehension　of　events　themselves,　encompassing　their　underlying　causes,　effects,　and　consequences.　This　helps　us　to　understand　and　explain　social　phenomena　in　a　more　comprehensive　way.　In　this　regard,　we　introduce　ChatGPT-opinion　mining　as　a　framework　that　transforms　event-centric　opinion　mining　tasks　into　question-answering　(QA)　utilizing　large　language　model.　We　employ　this　approach　in　the　context　of　the　event-centric　opinion　mining　task　that　utilizes　an　event-argument　structure.　In　our　study,　we　primarily　leverage　in-context　learning　methods　to　construct　demonstrations.　Through　comprehensive　comparative　experiments,　we　illustrate　that　the　model　achieves　superior　results　within　the　ChatGPT-opinion　mining　framework　when　the　demonstrations　exhibit　diversity　and　possess　higher　semantic　similarity,　comparable　to　those　of　supervised　models.　Moreover,　ChatGPT-opinion　mining　surpasses　the　supervised　model,　particularly　when　there　is　limited　availability　of　the　same　data.　©　2023,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.