bhanubathini2002/dataset_parsecdoc.ai

ParsecDoc LaTeX Templates

A collection of 4,186 LaTeX document templates, organized by document type. Each template folder contains:

• a compiled PDF (<template-name>.pdf) — the rendered document
• a source ZIP (<template-name>.zip) — the LaTeX source files used to produce it

This pairing of rendered output with its exact source makes the dataset useful for document parsing, PDF-to-LaTeX conversion, document layout analysis, and OCR training/evaluation.

Categories

Category	Templates
Theses	1,735
Presentations	738
Journal articles	588
Assignments	349
CVs and resumes	309
Posters	192
Formal letters	150
Books	93
Newsletters	14
Bibliographies	11
Calendars	7
Total	4,186

Structure

<Category>/
  <template-name>/
    <template-name>.pdf   # compiled document
    <template-name>.zip   # LaTeX source

The dataset viewer above shows the PDFs (the default config loads **/*.pdf as a pdffolder dataset). The ZIP sources are not part of the viewer config but can be downloaded directly (see below).

Usage

Load the PDFs with 🤗 Datasets

# requires: pip install "datasets[pdf]"  (installs pdfplumber)
from datasets import load_dataset

ds = load_dataset("bhanubathini2002/dataset_parsecdoc.ai", split="train")
print(ds[0]["pdf"])  # pdfplumber PDF object

Download the LaTeX sources (ZIPs)

from huggingface_hub import snapshot_download

path = snapshot_download(
    "bhanubathini2002/dataset_parsecdoc.ai",
    repo_type="dataset",
    allow_patterns="*.zip",
)

Download a single template

from huggingface_hub import hf_hub_download

pdf = hf_hub_download(
    "bhanubathini2002/dataset_parsecdoc.ai",
    "Assignments/assignment-template/assignment-template.pdf",
    repo_type="dataset",
)

Licensing

The templates were collected from publicly available template galleries. Each template retains the license chosen by its original author (commonly LPPL, MIT, CC-BY, or similar); check the source ZIP of an individual template for its specific license before redistribution.