So my transliteration tool is pretty good at finding high-probability letter pairs.

To use the case of Greek again, we get:

<h4>Current transliteration scheme</h4>

<pre><code>! a A ā ã ä b B c C d D e E é É ê f F g G h H I i í J k K l L m M n N o O ó p P q Q r R S s t T u U v V W w x Y y z Z
! α Α α ά έ β Μ κ Κ δ Δ ε Ε έ Έ λ φ Φ γ Γ α Χ Ι ί ί J κ Κ λ Λ μ Μ ν Ν ο Ο ο π Π κ κ ρ Ρ Σ ς τ Τ ο Ο β Β Λ ύ ξ Έ ύ ζ ζ
</code></pre>

It's all Greek to me, but it _seems_ like I could transliterate a lot of algebra equations with that mapping... but are all the letters there? 

Well, let's look at _all_ the letters on the Greek side of our lexicon, and compare. 
 
get_alphabet.py

<pre><code> ! " ' ( ) , - . / : ? A B C D E F G H I J K L M N O P Q R S T U V W X Y Z a b c d e f g h i j k l m n o p q r s t u v w x y z | µ ö ΄ Ά Έ Ή Ί Ό Ύ Ώ ΐ Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω ά έ ή ί ΰ α β γ δ ε ζ η θ ι κ λ μ ν ξ ο π ρ ς σ τ υ φ χ ψ ω ϊ ϋ ό ύ ώ
</code></pre>

Well, that's sort of hard to evaluate. For one thing, it's pretty clear that all the English letters also showed up on the Greek side. Presumably borrowings or acronyms or brand names or something. There are also clearly a bunch of accented letters like Ά, Έ, Ή. (You might need a special font to see those...)

There's simply no way to deal with this sort of thing automatically; because you end up with pairs like:

<pre><code>GNU Free Documentation License  Άδεια Ελεύθερης Τεκμηρίωσης GNU
</code></pre>

If we're collecting all the letters on the Greek side, we have to somehow "know" that "GNU" corresponds to "GNU." One hack around this would be to create a special sublist of pairs, where none of the words on the left appear on the right; or even where none of the letters on the left appear on the right. 


<pre><code>get_alphabet-2.py</code></pre>

<h4>All letters on Greek side of lexicon (more or less) minus Latin letters</h4>

<pre><code>( ) - . M X v ΄ Ά Έ Ή Ί Ό Ύ Ώ ΐ Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω ά έ ή ί ΰ α β γ δ ε ζ η θ ι κ λ μ ν ξ ο π ρ ς σ τ υ φ χ ψ ω ϊ ϋ ό ύ ώ </code></pre>

Eh, I dunno, looks Greekier.

Arrite, so, here's the point: which of those letters is not covered by the transliteration schema our approach has generated? 

<pre><code>find_missing.py</code></pre>

So now we need to find the difference of