It would not be the touch screen, even if it was an LCD or the Operating System at the time which would be expensive. Instead, it would be the cost of hire software developers to design a voice recognition for US/UK/AU/NZ English, German, Spanish and Japanese just to name the commercially dominating languages at the time. You also have to take into account all the voice tracks stored on it which would have to be in close to lossless since the MIDI, MOD and other audio compression at the time can't play voice tracks clearly. So quite a bit more is needed to create hardware with enough flash memory to store all that since hardrives and optical discs with sensitive data back then would have been unreadable if shaken while the car was riding on a bumpy surface.
With all that Subaru would have had to push up the the SVX to Lamborghini or Bentley pricing. It also would have been slow by most people's standards in processing voice to commands and become a major drain on the battery. So the point of it was to merely attract more people to it while at the autoshow or reading a magazine.
