STEPS TO REPRODUCE:
1. Open the attached pdf
2. Go to page 7
3. Highlight the fractions after question 11
4. Listen to the speech

ACTUAL RESULTS:
"nine, six, eighteen, ten, fifteen" are read before "twelve, eight, twenty four, sixteen, twenty".

EXPECTED RESULTS:
It should read "nine over twelve, six over eight, eighteen over twenty four, ten over sixteen, fifteen over twenty".

ADDITIONAL INFORMATION:
Alt text is present on the spread sheet which should overwrite the way the text is spoken.

Further information from our developer:
To speak selected text, we use `DocumentViewController.getSelectedText()`. But for words that have "<rdf:Alt>" property (in "<?xpacket ...><x:xmpmeta...>") we would like to speak user-defined alternative text instead of the printed one.
Is there any way to extract metadata for `PageText`, `LineText`, `WordText` and `GlyphText`?
If not, could you please add `getMetadata()` method to those classes so that we can access "<?xpacket>" dictionaries? Even binary `byte[]` result works for us, we will parse XML on our side.