README.md

metadata

dataset_info:
  features:
    - name: date_created
      dtype: timestamp[ns]
    - name: title
      dtype: string
    - name: categories
      dtype: string
    - name: arxiv_id
      dtype: string
    - name: year
      dtype: int32
    - name: embedding
      sequence: float64
    - name: data_map
      sequence: float64
  splits:
    - name: train
      num_bytes: 53050705
      num_examples: 8457
  download_size: 43069584
  dataset_size: 53050705
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Dataset Card for "arxiv_ml"

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
  • Supported Tasks and Leaderboards
  • Languages
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Dataset Creation
  • Curation Rationale
  • Source Data
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Dataset Curators
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

Dataset Summary

This is a dataset of titles and abstracts of category theory related papers from ArXiv. This data is derived from the ArXiv dataset available on Kaggle. The selection of papers was determined by selecting all papers that used a category tag "math.CT". To supplement the titles and abstracts the creation time of the paper, as well as the categories are provided. To make exploration easier embeddings of the title and abstract have been made using the Nomic-embed-v1.5 text embedding model, and a 2D representation using UMAP is also provided.

Supported Tasks

This dataset is primarily aimed at tasks such as topic modelling, corpus triage, search and information retrieval, and other NLP tasks.

Languages

The dataset is in English, although other languages may also be present.

Dataset Creation

Curation Rationale

The full ArXiv dataset is too large for many tasks. Subsetting to a selection of ArXiv categories related the category theory ensures a small sized dataset that should mostly contain topics that are familiar to those wishing to use the dataset.

Source Data

This data is derived from the ArXiv dataset available on Kaggle.

Personal and Sensitive Information

This dataset contains publicly published information that was available under a CC0: public domain license via Kaggle. There should be no personal or senstive information in this dataset. If this is in error, please contact the maintainer and we will endeavour to remedy any issues.

Additional Information

Dataset Curators

Leland McInnes for the curated subset, Cornell University for the initial full dataset.

Licensing Information

Licensed as CC0: Public Domain.