Recherches récentes
Options de recherche
Since #ChatGPT already does a decent job of postcorrecting #OCR errors, it would be interesting to see what it can achieve with fine-tuning now that it is available. #DigitizedNewspaper @cneud https://platform.openai.com/docs/guides/fine-tuning
platform.openai.comOpenAI PlatformExplore developer resources, tutorials, API docs, and dynamic examples to get the most out of OpenAI's platform.
Fine-tuning could be made with some ground truth data where the original OCR is available. I guess we should add the original OCR for our #OpenData on @huggingface @danielvanstrien https://huggingface.co/datasets/biglam/bnl_ground_truth_newspapers_before_1878
huggingface.cobiglam/bnl_ground_truth_newspapers_before_1878 · Datasets at Hugging FaceWe’re on a journey to advance and democratize artificial intelligence through open source and open science.