samchain/BIS_speeches_97_23_MLM
Dataset Card for "BIS_Speeches_97_23" This dataset is built from scrapped speeches on the Bank of International Settlements thanks to this repo : https://github.com/HanssonMagnus/scrape_bis. The dataset is made of 12k speeches from 1997 to 2023. Each pair is built with extracted sentences from speeches, if B is following A then the 'next_sentence_label' is 1 else it is 0. Negative pairs are built by choosing a sentence from another speech randomly. Credits Full… See the full description on the dataset page: https://huggingface.co/datasets/samchain/BIS_speeches_97_23_MLM.
Dataset Card for "BISSpeeches97_23"
This dataset is built from scrapped speeches on the Bank of International Settlements thanks to this repo : https://github.com/HanssonMagnus/scrapebis. The dataset is made of 12k speeches from 1997 to 2023. Each pair is built with extracted sentences from speeches, if B is following A then the 'nextsentence_label' is 1 else it is 0. Negative pairs are built by choosing a sentence from another speech randomly.
Credits
Full credits to the Bank of International Settlements for making the resource available through the internet. Credits to Samuel Chaineau for making it available on HF Hub.