Contents

At LokiBots, we are breaking the grounds of Robotic Process Automation and Conversational AI. And when we combine them together, we find ourselves to be in the brink of giving birth to a new civilization of service robots. So we recommend you read further on to find out how.

LokiBots Intelligence

LokiBots currently offers three main areas of intelligence: Robotic Process Automation, Document AI and Conversational AI.

Robotic Process Automation is done using our vision-based tool that can learn from video recordings of software and/or web operation.

Document AI is done using state-of-the-art OCR models along with personalization on client use cases.

Conversational AI is also done using state-of-art models, but of NLP/NLU, and with personalization on client use cases as well.

So now let’s take a look at how our Conversational AI developed from some of the gaps of the industries leading development frameworks, particularly Rasa and DialogFlow.

‍

CLICK HERE TO READ THE FULL CONTENT OF THIS BLOG ON MEDIUM

‍

Where are we heading now?

Much of our hard work and efforts have paid off really well when we designed our own Conversational AI tool, and we are able to design bespoke Conversational AI bots for many of our clients. But just as we design conversational bots for them, we do learn more in the process, and our tool gets updated with more new features and capabilities.

While our tool is great and something I would personally recommend you to try for your company, there are a few more gaps we are also looking to solve. The most important of all is identifying the conversation context. This is because, as much as we humans inherently like to be very verbose in our conversations, we also on the other hand like to be quite minimal in what we convey. And just like our sun deities of complex connections, we tend to use words (or determiners) like “this”, “that”, “those”, “these” and “it” to connect to earlier parts of our conversations or none at all (remember how the student kept saying “in the u.s.” in the demo video… that isn’t normal). And the way we humans identify what the correct connections are by what follows “those” words… or we just fill in the blanks if there’s none. Its only with the recent advancements with transformer models that we are able to provide the same for our bots. But then, there is still a lot of work required on how we can integrate such a module onto our existing framework.

When this is done, we would be taking one more step towards our goal of building our robotic civilization. Because as LokiBots says:

“Let humans do what they can do best, for everything else there is a bot.”

‍