I think the world needs an open source virtual assistant/vocal recognition framework. Why wait for Apple, Google and Microsoft to open up their service to third parties? The feature I want to see most on my digital device is a powerful virtual assistant. The problem is the complexity of building a virtual assistant/accurate vocal recognition. It involves fields of software that require very smart people and a lot of resources. Rather than list out what an open source virtual assistant would need to be possible, I'm going to talk about what would be possible if it was built.
The Best Way To Do It
Personally, I believe in the power of building a framework that third parties can add to. That is why I think OVA (Open Virtual Assistant) should be an app you download that allows you to download third party extensions to enhance. Forking it would be possible, but discouraged for all but the most niche uses. (Company assistant for businesses.) Third parties have an Extension Store they can sell, or give away their extensions in, all grouped together. Available for desktop platforms and Android because it would need deep access to other apps, and it needs to be launched from anywhere.
Launching OVA could be done with a keyboard shortcut or a voice command. It brings up a box for entering commands and showing data. For those who are more interested in vocal input it could be possible to turn off this box when launching with a voice command.
You would need to be able to designate default apps, so when you say,"play Music" it knows to pay from Spotify. (This would mean Spotify had to make an extension.)
App Specific Commands
Apps could build extensions that only work in their respective apps. Or developers could build services that only work in certain types of apps. For example, Ginger is a service for analyzing your writing and correcting your grammar. You could be writing in Ulysses and say, "Hey OVA check my grammar" and it uses Gingers servers to analyze your writing and show the grammatical errors.
It Doesn't Even Have To Use Artificial Intelligence
As long as it could understand a user and developers built in comprehensive commands it could get away with not having any virtual intelligence.
This will be a gargantuan task but it could have a bigger effect on computing than anything built by Apple or Microsoft. The problem is that if they did open Siri or Cortana to third parties it would make a service like this much less appealing for developers. Maybe it could find a home on Linux? I don't know, all I know is I want a powerful virtual assistant and I'm getting tired of waiting.