Kyber

Kyber refers to a group of NLP data management technologies associated with Stanny. The Kyber database is a MongoDB NoSQL database that stores various intents, examples, and responses. The Kyber application provides a comprehensive graphical interface to make data manipulation extremely easy.

Show me somethin'

Stanny & Kyber

Stanny is a complex NLP algorithm that would not exist without the extensive amount of data and data management capabilities Kyber provides. Stanny accesses Kyber through HTTP requests and interprets the JSON response to create sets of training data. Stanny does not manipulate the Kyber database but instead uses a snapshot to improve itself.

Previous Iterations

The original versions of Kyber were much different (and less useful) than the current variant. Version 1 was a series of Microsoft Excel spreadsheets with a simple data structure. When that became too complicated to manage, I moved the contents to a JSON file with the current data structure. I then moved the data to a MongoDB database to create the remote Kyber application.

Data Manipulation

Concise data manipulation is the sole purpose of the Kyber application. It exists as both a mobile and desktop application to make it easy to change elements anywhere. While you cannot rebuild the Stanny NLP model from a phone, you can change the training data away from a computer and create the new model later.

Data Structure

The main section of the Kyber database consists of "Capsule Stacks," a set of 4 consecutive layers that describe distinct areas of NLP intents. The first layer groups similar intents together, the second stores general intent information, the third contains separated lists of training data, and the fourth describes variable parts of training phrases, like the location present in "What is the weather in San Francisco?"

Have questions about Kyber?

Fill out my contact form with your question or comment and I'll get back to you as soon as I can!

Contact me