how to use skrutable
First thing: Try starting with the
dharmakṣetre example (opens in new tab). This sets Gītā 1.1 as the input in IAST Romanization and transliterates it to Devanāgarī.
For different transliteration output, use the "Output scheme" dropdown
and click "Transliterate" again.
To learn about supported transliteration schemes, visit the
skrutable manual on Github.
You can also use the "Swap" buttons to move things around as needed.
Now click between the "Transliterate", "Scan", "Identify Meter", and "Split Words" buttons to calculate and display skrutable's four basic outputs.
Scansion is line-by-line, and it doesn't matter how many lines or how much content you enter.
This mode is especially good for fiddling and exploring.
Meter identification, by contrast, effectively requires four lines of content,
corresponding to four verse quarters or pādas.
You should use this mode when you believe you're dealing with a whole verse
(half an anuṣṭubh śloka works, too).
You can use the "Re-Split Pādas" dropbox to decide whether to use your own line breaks
or to let skrutable try and find them for you.
The option "resplit_max" uses the most brute force and requires the least thought,
while "resplit_lite" better respects whatever line breaks (or ';' or ' / ') you provide.
Or, choose "none" in order to evaluate exactly the (necessarily four!) lines you input.
Definitely also try changing the mix-and-match scansion detail options with the checkboxes below "Scan".
These also work for the whole-file mode described below.
To learn more about working with Sanskrit meter, the
skrutable manual on GitHub provides some very basic pointers.
Finally, you can do sandhi and compound splitting using neural-net ML models trained on the Digital Corpus of Sanskrit.
(This is the work of Oliver Hellwig and Sebastian Nehrdich, who have kindly made their models available for use here.
See 2018,
and 2024 (publications forthcoming) to learn more about this work of theirs.)
Skrutable allows you to input Sanskrit text of virtually any length, with punctuation,
and it will return a very handy plain-text output, including with hyphens to show compounds (2024 model only).
As with metrical functions, transliteration is also taken care of for you, too.
The results are not always 100% correct, but I think you'll see, it's quite useful nonetheless!
whole-file processing
Skrutable's input window can handle a lot, up to 64mb, but if you're working with large amounts of text,
you may prefer to upload your input as a file instead.
To do this for transliteration, just set your input and output schemes as normal,
but then instead of hitting the "Transliterate" button, hit the "whole file" button underneath.
Continue through the next screen to upload your file,
then look for the output in your usual download location.
Whole-file word splitting is exactly as easy. Just choose your desired settings, then using the "whole file" button.
Note, though, that processing is considerably slower with the more accurate 2024 model than with the older 2018 one.
Doing whole-file metrical analysis is also somewhat similar, but the input must be more carefully prepared,
with exactly one verse per line and no extraneous material.
See the "Scan GRETIL" page to get a better idea of what this entails.
Some Bhagavadgītā files are provided below as good starter examples: