Datasets:

Modalities:
Tabular
Formats:
parquet
DOI:
Libraries:
Datasets
Dask
License:
anki-revlogs-10k / README.md
JarrettYe's picture
specify the path pattern
16df9da verified
metadata
license: other
license_name: anki-revlogs-10k
license_link: LICENSE
dataset_info:
  - config_name: revlogs
    splits:
      - name: all
        num_examples: 727738650
  - config_name: cards
    splits:
      - name: all
        num_examples: 208432491
  - config_name: decks
    splits:
      - name: all
        num_examples: 2066838
configs:
  - config_name: revlogs
    data_files:
      - split: all
        path: revlogs/user_id=*/data.parquet
  - config_name: cards
    data_files:
      - split: all
        path: cards/user_id=*/data.parquet
  - config_name: decks
    data_files:
      - split: all
        path: decks/user_id=*/data.parquet
tags:
  - memory
  - spaced repetition
  - knowledge tracing
pretty_name: Anki Revlogs 10K
size_categories:
  - 100M<n<1B

Introduction

Anki Revlogs 10K is a dataset of 10k collections from Anki for FSRS project. It is a random sample of collections with 5000+ revlog entries, so it should contain a mix of older (still active) users, and newer users.

The dataset contains three parts: revlogs, cards, and decks.

Revlogs

This dataset contains flashcard review records with the following fields:

  • card_id: Unique identifier for each flashcard
  • day_offset: Number of days since the start of user's first review
  • rating: User's rating of their recall performance (1: again, 2: hard, 3: good, 4: easy)
  • state: Card's learning state (0: new, 1: learning, 2: review, 3: relearning)
  • duration: Time spent on the card review in milliseconds (range: 0-60000)
  • elapsed_days: Days since the last review (-1 for new cards)
  • elapsed_seconds: Seconds since the last review (-1 for new cards)

The records have been pre-sorted in chronological order.

Cards

This dataset contains card-note-deck relationship mappings with the following fields:

  • card_id: Unique identifier for each flashcard
  • note_id: Unique identifier for each note, where a note can contain multiple cards
  • deck_id: Unique identifier for each deck, representing different study categories or subjects

Decks

This dataset contains deck hierarchy information with the following fields:

  • deck_id: Unique identifier for each deck
  • parent_id: ID of the parent deck, indicating the deck's position in the hierarchy
  • preset_id: ID of the preset configuration applied to the deck

License

This dataset is released under the anki-revlogs-10k License.

Related Projects

Anki Revlogs Dataset Builder

SRS Benchmark