The oddities of the bigram "ed"

[Jorge_Stolfi]

my question about "eed" was related to the suggestion that "ed" might be a single unit. While this is tempting, this then leaves the question whether "eed" is another unit, or should be read as "e" + "ed". Neither seem satisfactory to me. (words ending -edy can also end -ey, or -eedy, or -eey)

Rene, you know my view.  My proposed "true alphabet" has the following elements:

  12443.625000  0.09984 {a}
  21543.125000  0.17285 {o}
  15120.250000  0.12132 {y}

  5192.000000   0.04166 {q}

  11388.500000  0.09137 {d}
   9220.250000  0.07398 {l}
   5792.375000  0.04647 {r}
   2081.750000  0.01670 {s}

   5664.500000  0.04545 {ch}   3906.250000  0.03134 {che}
   3842.250000  0.03083 {ee}    324.000000  0.00260 {eee}
   2085.500000  0.01673 {sh}   1869.250000  0.01500 {she}

   7295.000000  0.05853 {k}    1519.000000  0.01219 {ke}
   4173.750000  0.03349 {t}     786.000000  0.00631 {te}

    452.500000  0.00363 {ckh}   169.500000  0.00136 {ckhe}
    591.000000  0.00474 {cth}   148.000000  0.00119 {cthe}

    316.000000  0.00254 {f}
   1197.375000  0.00961 {p}

     36.000000  0.00029 {cfh}
    104.500000  0.00084 {cph}

    113.500000  0.00091 {n}     868.250000  0.00697 {m} 
   1665.500000  0.01336 {in}
   3779.000000  0.03032 {iin}
    159.000000  0.00128 {iiin}
   
My You are not allowed to view links. Register or Login to view. listed a few others, like {cphe} and {ir}, but now I am almost convinced that they are errors at some stage. 
  
Thus, in my model, 

• A single e is always a modifier for the previous element, which must be a "bench" Ch, Sh, or a "kite" (k/t gallows) possibly with platform.
  
• A double e is always an element {ee} by itself, in the same "benches" class as Ch and Sh.
  
• A triple e can be an element {eee} by itself; but if it comes after a bench or kite, it is ambiguous. Like, Cheeed could be {ch}{eee}{d} or {che}{ee}{d}. One day we may know the rule, but for now either parsing may be the right one.
  
• A quadruple e may be ambiguous too, but they must be too rare to matter.
  

This model would "explain", for instance, why ed is never word-initial.  And it predicts that de is possible only as dee, not with a single e. Is this the case?

All the best, --stolfi