---
dataset_info:
  features:
  - name: sequenceA
    dtype: string
  - name: sequenceB
    dtype: string
  - name: next_sentence_label
    dtype: int64
  splits:
  - name: train
    num_bytes: 505762257.6721524
    num_examples: 773395
  - name: test
    num_bytes: 89252509.32784761
    num_examples: 136482
  download_size: 365034957
  dataset_size: 595014767
license: apache-2.0
task_categories:
- text-classification
- token-classification
language:
- en
tags:
- economics
- finance
- business
size_categories:
- 100K<n<1M
---
# 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.)