If a hyphenated word is stored as-is in the PDF output, a screen reader will read it differently to when it is not hyphenated. This can result into incomprehensible text. To fix that problem, a hyphenated word should properly be tagged as such. This can be done in 2 ways. The first possibility is to add an 'ActualText' entry to the property list of the corresponding marked-content sequence. Its value would basically be the whole text minus the last hyphen character. The second possibility is to replace the last hyphen with a soft hyphen character, which will be recognized by screen readers such that the split word will be read as one. This will work only if the font has a glyph for the soft hyphen character. The latter possibility is the recommended way to handle hyphenated words. The former can be implemented as a fallback for when there is no available glyph for the soft hyphen, or when the hyphenation character is not actually a hyphen (this can be customized through the hyphenation-character property).
Fixed in rev. 1405158: http://svn.apache.org/viewvc?rev=1405158&view=rev