Natural Language Theory for Mobile Apps. And why it should be Abstracted

I’m going to be honest – I’ve looked at a couple of my competitors’ solutions and I must say that the final result they’ve got in their demos looks really awesome. But like any great dish, the final product appears to rely a lot on the chef.

Now, if only I could understand how to use their many different objects, entities, actions, interactions and extensibility and whatnot. After playing around on their site for a little, I begin to wonder if there’s some kind of pre-usage workshop I could attend, because the YouTube ‘getting started’ video is clearly useful for few other applications than the one the demo was being put together for. I’m sure that their tech has a WHOLE LOT of applications, but that’d require me to learn their proprietary way of interacting with their proprietary objects. And that’s a whole lot of new stuff for apps to learn. But why?

While building Leova, we had one single consideration in mind - the awesome app developer out there. An app developer is one of the brightest bulbs in the box, and they do a TON of work to enhance and constantly upgrade their skillset. And it’s a lot of work, I’ve got to be honest (we know because we tried building a couple of apps and we suck at it)!

So while designing a workflow, we asked ourselves, “how do we allow app developers to build better apps with voice, without having to learn ANYTHING about Natural Language Processing (NLP), but continue to deliver outstanding experiences to their users?”’

Now that’s a tough question. And it’s one that many very smart people around the world are trying to solve. While most of the solutions are being crafted using big data, we took a different, higher-speed approach – we looked at how people speak to each other, in great detail. We studied the structure of conversations and what constituted a fully-formed conversation vis-à-vis conversations that appeared to be stilted, dysfunctional and just plain unnatural sounding. And this was hard: English is full of exceptions and variations from standard rules, especially in modern day usage.

While building Leova we took the hybrid approach of supplementing textbook natural language analysis, of Bayes and Markov, with technology that we built ourselves. Our focus has been on not just building technology that can understand human language, but also building an expertise-base (narrow AI of sorts) in certain specific sectors, to emulate the experience of talking to an expert in that field (much like having a conversation with an experienced travel agent). This explains why we’ve been staggering our releases – travel, next food and early next year, IoT.

If this sounds interesting, we urge you to come visit the Leova page: https://www.leova.io, it takes nothing more than 2 minutes to get an implementation up and running. And another minute or so to personalize the LeovaTravel implementation with destinations.