Skip to content

v1.0.1-gui

Latest
Compare
Choose a tag to compare
@hyunjongkimmath hyunjongkimmath released this 13 Jan 01:23

This is the first release of the GUI for trouver. The GUI has only a subset of the functionalities available to trouver:

image

  1. Download source .tex files from arXiv.org

    • If the GUI fails at this task for an arXiv article, it may be necessary download and decompress the arXiv source file manually.
  2. Divide a .tex file (not necessarily one from arXiv.org) into moderately-lengthed "parts" by creating an Obsidian.md vault whose notes are made of these parts.

  3. Use the ML models hyunjongkimmath/def_and_notat_token_classification_model and hyunjongkimmath/information_note_type to identify which parts introduce definitions and notations and to find where the definitions and notations are introduced at

  4. Compile the parts back into a single .tex file, effectively annotating the original .tex file with the locations of the definitions and notations introduced --- When this annotated .tex file is built, the locations where definitions and notations are introduced are highlighted in light grey as below:

image

(the above is an excerpt from Hyun Jong Kim and Sun Woo Park's joint paper [2106.10586] Global $\mathbb{A}^1$ degrees of covering maps between modular curves)

However, note that the ML models are prone to error. It may also be necessary to manually modify the .tex to fix syntax errors that may arise during the annotation process.

The GUI application will likely take significantly longer than a few seconds to open.

The GUI application downloads the relevant ML models from huggingface hub to your local machine. Afterwards, model predictions are performed on your local machine. The GUI application does NOT collect your data.

Due to the variety of ways that mathematical writers format their .tex files, this application may not work well on certain .tex files. trouver and this GUI application are not tested extensively on MacOS, and hence there may be issues when running on MacOS. Please consider reporting issues to the issues page of this repo.

Please consider using trouver as a Python library/pip package as it has more features, is better supported, and may be more stable than the GUI application. Please also consider filling out the following survey.