Speech to Text

Means for PA announcements to be digitized automatically from the spoken language.  English is presumed as the basic language.  The English text could be translated into other languages or forms of conveyance.

A trained human can listen to speech and transcribe text in real-time with a high degree of success.  Significant industry attention is underway to progress voice commands as applicable to home networking attendants, home media control, and mobile device interaction.  The objective is to be able to attach to an audio source and in real-time, create an English language (others to follow) text recognition, whether with access to remote server processing or whether operating in a local mode only. 

Airlines will be encouraged to solicit audio recordings from passenger address systems to build up a representative set of reference data and reserved data.  The reference data will be released to any interested party where they are encouraged to submit their text results against a scoring sheet.  The reserved data will be used by WiFIND.org to evaluate any promising candidate solutions.  Airlines will then be solicited to participate in any trials where selected solutions will be installed for limited flight testing.