metadata

dataset_info:
  - config_name: documents
    features:
      - name: chunk_id
        dtype: string
      - name: chunk
        dtype: string
    splits:
      - name: train
        num_bytes: 3968308
        num_examples: 5502
      - name: validation
        num_bytes: 399556
        num_examples: 555
      - name: test
        num_bytes: 1263082
        num_examples: 1750
    download_size: 3462955
    dataset_size: 5630946
  - config_name: queries
    features:
      - name: og_query
        dtype: string
      - name: query
        dtype: string
      - name: chunk_id
        dtype: string
      - name: answer
        dtype: string
    splits:
      - name: train
        num_bytes: 6371715.290749214
        num_examples: 27953
      - name: validation
        num_bytes: 664723
        num_examples: 2919
      - name: test
        num_bytes: 1970467.8313323231
        num_examples: 8575
    download_size: 3619450
    dataset_size: 9006906.122081537
configs:
  - config_name: documents
    data_files:
      - split: train
        path: documents/train-*
      - split: validation
        path: documents/validation-*
      - split: test
        path: documents/test-*
  - config_name: queries
    data_files:
      - split: train
        path: queries/train-*
      - split: validation
        path: queries/validation-*
      - split: test
        path: queries/test-*

ConTEB - NarrativeQA

This dataset is part of ConTEB (Context-aware Text Embedding Benchmark), designed for evaluating contextual embedding model capabilities. It stems from the widely used NarrativeQA dataset.

Dataset Summary

NarrativeQA (literature), consists of long documents, associated to existing sets of question-answer pairs. To build the corpus, we start from the pre-existing collection documents, extract the text, and chunk them (using LangChain's RecursiveCharacterSplitter with a threshold of 1000 characters). Since chunking is done a posteriori without considering the questions, chunks are not always self-contained and eliciting document-wide context can help build meaningful representations. We use GPT-4o to annotate which chunk, among the gold document, best contains information needed to answer the query.

This dataset provides a focused benchmark for contextualized embeddings. It includes a set of original documents, chunks stemming from them, and queries.

Number of Documents: 355
Number of Chunks: 1750
Number of Queries: 8575
Average Number of Tokens per Chunk: 151.9

Dataset Structure (Hugging Face Datasets)

The dataset is structured into the following columns:

documents: Contains chunk information:
- "chunk_id": The ID of the chunk, of the form doc-id_chunk-id, where doc-id is the ID of the original document and chunk-id is the position of the chunk within that document.
- "chunk": The text of the chunk
queries: Contains query information:
- "query": The text of the query.
- "answer": The answer relevant to the query, from the original dataset.
- "chunk_id": The ID of the chunk that the query is related to, of the form doc-id_chunk-id, where doc-id is the ID of the original document and chunk-id is the position of the chunk within that document.

Usage

Use the train split for training, and the test split for evaluation. We will upload a Quickstart evaluation snippet soon.

Citation

We will add the corresponding citation soon.

Acknowledgments

This work is partially supported by ILLUIN Technology, and by a grant from ANRT France.

Copyright

All rights are reserved to the original authors of the documents.