The package eSpeak is a compact open source software speech synthesizer for English and other languages and it is available in AlexT's repo, you will find easy instructions on how to download from the repo here: Edison Package Repo Configuration Instructions.
If you need further information about eSpeak, I suggest you to visit eSpeak: Speech Synthesizer.
I too was looking for a good Voice Recognition library suitable for Edison (voice-to-text not text-to-voice like espeak)
I have got the following working
1. Using the google cloud voice to text service
Very accurate and fast - but it requires an internet connection and I was looking to use if offline.
2. I got pocketsphinx working...but
Issue is its quite slow but main issue is its very very inaccurate getting most words close but wrong unless you severely limit its word list.
There was an offline one promised several years ago at the Edison launch - Jarvis (link: Intel Jarvis Voice Recognition without Going Cloud - Bye Bye Siri? ) - was this library ever made available to developers?