I prepared a WWW page with the word pair tables above.
  
  I also prepared analogous tables for the English and Portuguese texts:
  
    cat engl.txt | sed -e 's@$@ //@g' | tr ' ' '\012' | egrep '.' > engl2.wds
    
    cat engl2-keys.dic
    
      //
      the
      a
      an
      and
      of
      in
      on
      at
      to
      for
      with
      as
      up
      but
      i
      you
      he
      she
      it
      is
      was
      had
      be
      my
      his
      her
      him
      me
      that
      mrs
      john
      cynthia
      inglethorp

    cat engl2.wds | tr '[A-Z]' '[a-z]' | head -4661 \
      | enum-word-pairs \
      | count-diword-freqs -v keyfile=engl2-keys.dic \
      > .baz
       
    cat port.txt | sed -e 's@$@ //@g' | tr ' ' '\012' | egrep '.' > port2.wds
     
    cat port2-keys.dic
    
      //
      a
      da
      na
      o
      do
      no
      ao
      as
      das
      os
      dos
      um
      uma
      cada
      de
      em
      por
      para
      e
      ou
      como
      que
      é
      ser
      não
      são
      aresta
      face
      arestas
      faces
      complexo
      vértices
      celular

    cat port2.wds | sed -e 's/^x$/???/g' | head -7000 \
      | enum-word-pairs \
      | count-diword-freqs -v keyfile=port2-keys.dic \
      > .baz 
      
  The results were posted on my Voynich WWW page.
  
  Decided to recompute the tables, adding the left and right probabilities. 
  
     cat .wds \
      | sed -e '/?/s/^.*$/???/g' \
      | enum-word-pairs \
      | count-diword-freqs -v keyfile=.keys \
      > .baz