JohannesBuchner / languagecheck

Language checking for scientific papers

This program attempts to assist you in improving your paper before submission.

Features

• Can analyse any LaTeX papers, and Overleaf projects.
• Makes automated reports to point you to improvements:
  • Word level:
    • find common grammar mistakes, like wrong prepositions
    • find wordy phrases and suggest replacements
    • a vs an
    • spell-check (using hunspell)
  • Sentence level:
    • find long, wordy sentences
    • check topic sentences
  • Paragraph level:
    • find tense inconsistencies
  • Paper level:
    • check visual impression of paper
• All analysis is done offline -- your text does not leave your computer.
• Supports British and American English, but focusses on issues applying to both.

Note that there are false positives -- only you can decide whether a change would make sense, the reports only point out potential issues.

If you find some rules useless (too many false positives), or you want to add more, please send a pull request!

Demo output

Example analysis (of an early draft of this paper):

• Example report for misused phrases
• Overview of all reports

Requirements

• python
• convert command (ImageMagick): Install with your distribution
• nltk: Install with pip
• nltk data: Install with python -m nltk.downloader all
• detex command (usually comes with LaTeX)
• pyhunspell (optional): Install with pip

Installation

These commands should not give you an error:

$ which convert
$ which python
$ which detex
$ which hunspell
$ ls /usr/share/hunspell/{en_US,en_UK}.{dic,aff}

Then install the python packages and data:

$ pip install pyhunspell  --user
$ pip install nltk  --user
$ python -m nltk.downloader cmudict stopwords punkt averaged_perceptron_tagger

Usage

Using directly:

• create PDF from your latex file -> mypaper.pdf
  • For example, run "pdflatex mypaper.tex"
• use detex to create pure text file -> mypaper.txt
  • For example, run "detex mypaper.tex > mypaper.txt". You need detex installed.
  • This does not capture figure captions. The detex.sh script can help you include those texts, "bash detex.sh mypaper.tex". You still need detex installed
• run $ python languagecheck.py mydir/mypaper.txt mydir/mypaper.pdf
• open with a web browser mypaper_index.html to see all reports

Using with Overleaf:

$ bash languagecheck_overleaf.sh <overleaf_url> <name of tex file>
# for example:
$ bash languagecheck_overleaf.sh https://www.overleaf.com/123456789 mypaper.tex

See also

• style-check, a similar program written in Ruby: https://github.com/nspring/style-check/

• Statistics checklist: Check for common statistics mistakes with this checklist

  http://astrost.at/istics/minimal-statistics-checklist.html