Inventing the Voice Assistant User Interface (Part 2)

Inventing the Voice Assistant User Interface (Part 2)

Voice Assistant (VA) developers are increasingly under pressure to create “naturalistic voice experiences” despite the current limitations of Voice Recognition (VR) and Voice Parsing (VP) technology. Instead we as developers have to invent new tools and approaches to compensate for the technology’s limitations.

One approach is teach users to speak in ways that computers understand and computers to understand human beings more fluently. The first use case example is the Method of Loci. For more on the previous discussion, please see Inventing the Voice Assistant User Interface (Part 1).

The second approach also has historical roots that arose due to the limits of technology – LambdaMOO.

Example #2: LambdoMOO

LambdoMoo is an old text adventure game harness that incorporates constrained syntax. All commands are based in spoken English:

Verb,

Verb/Direct-Object or

Verb/Direct-Object/Preposition/Indirect Object.

It is a limited framework that still provides scope for complexity. Programmers could teach users a variety of commands to give a sense of operational control to users. Users could navigate the gaming scenarios easily after only brief tutorials. The challenge is in the discovery of additional commands to enrich user experience.

(In this sense LambdaMOO is less intuitive than the Method of Loci approach. Both approach are still superior to fully open text at this time.)

Best of all, LambdoMOO is sufficiently constrained to provide well-defined decision points. The computer has a limited number of options to consider to work out what the user says, which lessens the reliance on VR/VP accuracy/access. Developers can include more intelligent prompting as well.

* * *

With Method of Loci as well as LambdaMOO, we are deliberating trading naturalistic speech for more efficient communication. They are compromises built on the current limits of VR/VP technology (much of which is beyond developers’ control). These approaches make sense in situations where the goal is to accomplish tasks rather than convey feeling or emotion. They are not tools of conversation or relationship-building. They are the vocal equivalent of joysticks.

What we would like to see is the standardization of VA approaches, whatever their form. Such agreement would benefit the development and user communities by reducing learning curves. They would also free up the development community to focus on further improvements, which, again, have yet to be invented.

Posted in Alexa, Amazon, Google, Google Assistant, IT Solutions, Management Consulting, Voice Assistant