FemkeBakker's picture
Update README.md
aff4190 verified
metadata
language:
  - nl
license: eupl-1.1
size_categories:
  - 10K<n<100K
tags:
  - documents
  - fine-tuning
dataset_info:
  features:
    - name: prompt_id
      dtype: int64
    - name: message
      list:
        - name: content
          dtype: string
        - name: role
          dtype: string
  splits:
    - name: train
      num_bytes: 9011452
      num_examples: 9900
    - name: test
      num_bytes: 998068
      num_examples: 1100
    - name: val
      num_bytes: 1000675
      num_examples: 1100
    - name: discard
      num_bytes: 7897005
      num_examples: 8718
  download_size: 5846654
  dataset_size: 18907200
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
      - split: val
        path: data/val-*
      - split: discard
        path: data/discard-*

This dataset is a modified version of the AmsterdamDocClassificationDataset. The original dataset consists of Dutch Raadsinformatie documents from the Municipality of Amsterdam, which were published in accordance with the Open Government Act (Woo). In this modified version, the documents are truncated to the first 200 tokens each. This dataset is used to fine-tune large language models (LLMs) for the Assessing LLMs for Document Classification project. The documents are formatted into a zero-shot prompt and then turned into conversations, where the ideal response of the model is the prediction (class) formatted into JSON format.

Specifics:

  • Truncation: first 200 tokens of each document. Docs are tokenized using the Llama tokenizer.
  • Data split:
    • test set: first 100 docs of each class (in total 1100 docs)
    • train set: remaining docs, with max of 1500 docs per class (11000 docs)
      • train set: 90% of train set is used for fine-tuning model (9900 docs)
      • val set: 10% of train set is used for evaluating the loss during training (1100 docs)

Data sources:

This dataset is part of [inseart thesis info] in collaboration with Amsterdam Intelligence for the City of Amsterdam.