arxivparse

Convert arXiv papers to clean plain text using LaTeXML. Download any arXiv paper and get structured text with title, abstract, sections, and math notation preserved.

Install

pip install arxivparse

Quick verification that LaTeXML is installed:

latexml --version

Usage

Get text as a string

from arxivparse import arxiv_to_text

text = arxiv_to_text("1706.03762")  # Attention is All You Need paper
print(text[:200])

Save text to a file

from arxivparse.pipeline import convert_arxiv_to_text
from pathlib import Path

output_path = convert_arxiv_to_text("1706.03762", output_path="paper.txt")
text = Path("paper.txt").read_text()

Handle errors gracefully

from arxivparse.errors import (
    Arxiv2TextError,
    DownloadError,
    NoLatexSourceError,
    ConversionError,
)

try:
    text = arxiv_to_text("1706.03762")
except NoLatexSourceError:
    print("Paper is PDF-only")
except DownloadError:
    print("Download failed")
except ConversionError:
    print("LaTeXML conversion failed")

Library Usage

Simple: get text as a string

from arxivparse import arxiv_to_text

text = arxiv_to_text("1706.03762")

Full pipeline control

from arxivparse.pipeline import convert_arxiv_to_text
from arxivparse.errors import Arxiv2TextError

try:
    output_path = convert_arxiv_to_text("1706.03762")
    text = output_path.read_text(encoding="utf-8")
except Arxiv2TextError as e:
    print(f"Failed: {e}")

Configuration

Both functions accept optional timeouts:

# arxiv_to_text() - returns string
text = arxiv_to_text(
    arxiv_id="1706.03762",
    download_timeout=60,    # seconds to wait for download
    convert_timeout=120,   # seconds to wait for LaTeXML
)

# convert_arxiv_to_text() - returns file path
path = convert_arxiv_to_text(
    arxiv_id="1706.03762",
    output_path="paper.txt",  # custom output file
    download_timeout=60,
    convert_timeout=120,
    keep_temp=False,         # keep intermediate files
)

Error Handling

from arxivparse.errors import (
    DownloadError,         # network/HTTP failure
    NoLatexSourceError,    # paper is PDF-only
    ConversionError,       # LaTeXML failed
    MainTexNotFoundError,  # no .tex file found
    Arxiv2TextError,       # any pipeline failure
)

CLI Reference (Optional)

If you prefer the command line:

# Install the CLI (comes with the package)
arxivparse 1706.03762

# Multiple papers
arxivparse 1706.03762 2301.07041

# Custom output
arxivparse -o paper.txt 1706.03762

# Verbose output
arxivparse -v 1706.03762

# Keep temp files for debugging
arxivparse --keep-temp 1706.03762

Output Format

QASPER-style plain text: title, abstract, section headings, body paragraphs, inline math (LaTeX notation like h_{t}), and figure/table captions. No bibliography entries, footnotes, author affiliations, or citation numbers.