Skip to main content

Automated notion clustering for the knowledge LaTeX package

Project description

Knowledge-Clustering

PyPI

Clustering notions for the knowledge LaTeX package. Maintained by Rémi Morvan, Thomas Colcombet and Aliaume Lopez.

Principle

The goal of Knowledge-Clustering is, when using the knowledge package to automatically provide suggestions to the user of what notions should be grouped together.

Installation

To install (or upgrade) Knowledge-Clustering, run

python3 -m pip install --upgrade knowledge-clustering

and then

knowledge init

Syntax

Usage: knowledge cluster [OPTIONS]

  Edit a NOTION file using the knowledges present in a DIAGNOSE file.

Options:
  -n, --notion FILE         File containing the notions that are already
                            defined.  [required]
  -d, --diagnose FILE       Diagnose file produced by LaTeX.  [required]
  -l, --lang [en]           Language of your TeX document.
  --scope / -S, --no-scope  Print the scopes defined in the notion file and
                            print the possible meaning of those scope inferred
                            by Knowledge Clustering.
  -c, --config-file TEXT    Specific configuration file. By default the
                            following files is read
                            /Users/rmorvan/GDrive/Code/knowledge-
                            clustering/knowledge_clustering/data/english.ini
  --help                    Show this message and exit.

Example

Example files can be found in the examples/ folder.

While writing some document, you have defined some knowledges in a file called small.tex (distinct from your main LaTeX). You continued writing your LaTeX document (not provided in the examples/ folder) for some time, and used some knowledges that were undefined. When compiling, LaTeX and the knowledge package gives you a warning and writes in a .diagnose file some information explaining what went wrong. This .diagnose file contains a section called "Undefined knowledges" containing all knowledges used in your main LaTeX file but not defined in small.tex. We reproduced this section in the small.diagnose file.

Screenshot of the small.tex and small.diagnose files before running Knowledge-Clustering. small.tex contains four knowledges, while small.diagnose contains five undefined knowledges.

Normally, you would add every undefined knowledge, one after the other, in your small.tex. This is quite burdensome and can largely be automated: you don't need a PhD to understand that "word" and "words" are similar words. This is precisely what Knowledge-Clustering does: after running

knowledge cluster small.tex small.diagnose

your file small.diagnose is left unchanged but small.tex is updated with comments.

The cluster command is optional: you can also write knowledge small.tex small.diagnose.

After running Knowledge-Clustering, the five undefined knowledges are included in the small.tex file as comments.

Now you simply have to check that the recommandations of Knowledge-Clustering are correct, and uncomment those lines.

Devel using virtualenv

Using virtualenv and the --editable option from pip3 allows for an easy setup of a development environment that will match a future user install without the hassle.

For bash and Zsh users

virtualenv -p python3 kw-devel
source ./kw-devel/bin/activate
pip3 install --editable .

For fish users

virtualenv -p python3 kw-devel
source ./kw-devel/bin/activate.fish
pip3 install --editable .

FAQ

  • When running knowledge, I obtain a long message error indicating "Resource punkt not found."

    Solution: run knowledge init.

  • My shell doesn't autocomplete the command knowledge.

    Solution: depending on whether you use zsh or bash write

     eval "`pip completion --<shellname>`"
    

    (where <shellname> is either zsh or bash) in your .zshrc (or .bashrc) file and then, either lunch a new terminal or run source ~/.zshrc (or source ~/.bashrc).

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

knowledge-clustering-0.2.3.tar.gz (12.7 kB view hashes)

Uploaded Source

Built Distribution

knowledge_clustering-0.2.3-py3-none-any.whl (13.6 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page