mw72.txt deviations from AS

[drdhaval2785]

Uses 'sh' for 'S' r2i for 'f' c4 for 'c' n6 for 'M' n3 for 'Y' n4 for 'N' lr2i for 'x' l2 for 'L'

[funderburkjim]

The file which governs the conversion from AS to slp1 for the headwords in MW72 is as_slp1.xml.

At the bottom is a list of non-standard AS forms. Pretty much what you listed above.

<!-- mw72 modifications -->
<e><s>INIT</s><in>n3</in><out>Y</out></e>
<e><s>INIT</s><in>n4</in><out>N</out></e> <!-- in orig, n.¤ -->
<e><s>INIT</s><in>c4</in><out>c</out></e>
<e><s>INIT</s><in>c4h</in><out>C</out></e>
<e><s>INIT</s><in>m2</in><out>M</out></e>  <!-- not special to mw72 -->
<e><s>INIT</s><in>n6</in><out>M</out></e>

<e><s>INIT</s><in>r2i</in><out>f</out></e>
<e><s>INIT</s><in>r2i1</in><out>F</out></e>
<e><s>INIT</s><in>lr2i</in><out>x</out></e>
<e><s>INIT</s><in>lr2i1</in><out>X</out></e>
<e><s>INIT</s><in>sh</in><out>z</out></e>

<!-- Dec 16, 2015. -->
<e><s>INIT</s><in>l2</in><out>L</out></e>
<e><s>INIT</s><in>l2h</in><out>|</out></e>

You might find that similar files for other dictionaries are relevant for your current work.

One thing which we are now probably in a position to do is to do these odd conversions once and for all in the reference digitization (e.g. in mw72.txt), so we don't have to keep remembering these transient details.

[drdhaval2785]

Jim, I will hug you tight, if you can kill these differences once and for all. All Sanskrit in SLP1 please.

[gasyoun]

One thing which we are now probably in a position to do is to do these odd conversions once and for all in the reference digitization

Yes. Some details might get lost in UTF8 (as I remember after exploring this field in 2014), but the hell with them if they add so many issues.

[drdhaval2785]

https://github.com/sanskrit-lexicon/MW72/issues/3 killed the differences.