gtexfix

Fix for Google Translate to process LaTeX documents.

Description

Code to.py replaces the LaTeX constructs by tokens. After passing translation the tokens are then restored by from.py. Simple tokens of the type [number.number] are used, which are more friendly to Google Translate. If the token type conflicts with the original text the user is notified. At times, Google Translate will corrupt the tokens and may even change their numbers unpreditably (the side effect of machine learning!). Corrupted tokens are identified and reported to the user for manual treatment.

Usage

1. Run $./to.py <source.tex> to produce <source_*.txt>. Several files are produced if the output exceeds the Google Translate character limit.

2. Feed files <source_*.txt> to Google Translate and merge them to obtain <translation.txt>.

3. Run $./from.py <translation.txt> to produce <translation.tex>.

4. Check for the corrupted tokens. If needed, edit manually <translation.txt> and run step 3 again.

Example

Sample LaTeX input is give in the examples folder.

Step 1. Run to.py:

$ ./to.py examples/example.tex
LaTeX file: examples/example.tex
No token conflicts detected. Proceeding.
Output file: examples/example_0.txt
Supply the output file(s) to Google Translate

Step 2. Google Translate -> Choose document "example_0.txt". Save the output page as a txt document (e.g. translation.txt).

Step 3. Run from.py:

$ ./from.py examples/translation.txt 
Input file: examples/translation.txt
Output file: examples/translation.tex
Corrupted tokens detected: [1.57] 
To improve the output manually change the corrupted tokens in file examples/translation.txt and run from.py again.

Step 4. In a text editor fix the corrupted token [1.57] in translation.txt. Run from.py again:

$ ./from.py examples/translation.txt 
Input file: examples/translation.txt
Output file: examples/translation.tex
No corrupted tokens. The translation is ready.

Alternatives

There is an alternative tool GoogleTranslate4LyX (https://wiki.lyx.org/Tools/GoogleTranslate4LyX) which converts LaTeX to HTML and uses <span class="notranslate"> tag instead of tokens to pass the Google Translate safely. However, it does not treat math formulas.

Possibilities for extension

The token numbers can be unpredictably changed by Google Translate: e.g. 396 in the original can become 369 after translation for no apparent reason. Although, the corrupted tokens are reported, the notranslate tags in HTML could in principle give a more stable solution than the use of tokens.