ONDEWO Natural Language Understanding (NLU) Platform

You can understand the natural language and automate conversations with our high-performance software platform.

Benefits

What you gain

The ONDEWO Natural Language Understanding (NLU) Platform is a highly scalable, high-performance software platform for enterprises to understand natural language and automate human conversations. It combines algorithms which enable the AI to understand complex topics and improve over time, which permits it to handle independently more tasks. 

Customers use ONDEWO to add cutting-edge natural language understanding capabilities to various products and services (e.g., chatbots, voicebots, email routing, document analysis, intent detection, named entity recognition and many more). 

Important Features

What makes it special

Cutting-edge deep learning and machine learning algorithms from research

Our natural language processing, machine learning and deep learning algorithms are directly from our research, joint research with our collaborators or from papers of the best researchers in the world. We usually need 3-4 weeks to bring new algorithms, methods, best-practices and pre-trained models into our product and in production for our customers. This incredible speed of innovation catapults our customers at the forefront of natural language understanding.

Automation of many NLU tasks and processes

The following processes can get fully automated: entity recognition, intent detection, context understanding, answer generation, training data augmentation, conversation flow state tracking and context switch detection as well as our novel question answering algorithms allowing the AI to answer questions from crawling thousands of web pages without designing entities or intents.

Multi-tenant and multi-project

The platform is a fully featured multi-tenant solution capable of serving multiple customers with multi-projects at the same time while utilizing the same underlying hardware and data storage. All customer-specific data is securely separated and backup-able in different databases.

Multiple languages and multiple response channels per project

Within a project it is possible to design, train and maintain intents, entities, responses and conversation flows for multiple languages. Additionally responses can be designed for specific output channels (e.g. only text for telephone conversations or voice assistants, chat interactive for chatbot-like interfaces and robots for Pepper and NAO-like robots from Softbank Robotics) 

Identity access management with roles and detailed access rights

Detailed access rights and restrictions can be applied to every single endpoint of the platform assigned to groups of users (roles). The platform features standard roles, however also allows to create new roles with detailed access restrictions and right. 

Quick and easy integration into products and services

All platform capabilities are easy to integrate into your products and services via ondewo-client libraries for various programming languages (e.g. Python, Nodejs, Angular, JavaScript etc.) and gRPC Remote Procedure Calls (“GRRPC”).

On-premise and cloud hosting

All features are available on both versions. The on-premise solution easily installs on all operating systems with standard Docker deployment environment on your own IT Infrastructure. Furthermore, it allows for a higher degree of data protection and control. With the cloud option, there is no maintenance needed as it gets provided by ONDEWO. The cloud option is billed on a usage basis. 

OUR UNIQUE AI APPROACH

Distinct accuracy with our meta-learning algorithm approach

Understanding natural language is very difficult and a complex problem since there are many variations how humans can express their intents and what they want to convey. Competitors mostly use one or two algorithms to come to a conclusion which entities or which intent should be recognized and tune this manually. Some even only use one algorithm for entity recognition and one for intent detection. They are missing out on the opportunity to train an AI to find the best answer to a customer questions.

We developed an unique meta learning algorithm approach for natural language understanding to significantly improve the quality of the recognition results for entities and intents by combining over 20 algorithms from natural language processing (e.g. similarity and search algorithms), machine learning (e.g. fasttext and word embedding) and deep learning algorithms (e.g. CNN, DNN, LSTM, BERT, GPT2 and other transformer models)  and have an AI meta algorithm that learns fully automatically which algorithms to ask and the degree of trust in the respective algorithm’s answer for a specific customer request.

ARCHITECTURE

Best-practice NLU architecture and services

ONDEWO Natural Language Understanding (NLU) Platform implements a best practice architecture from Gartner and realizes the various NLU capabilities as microservices. These allows for fully distributed training of the more than 20 AI algorithms on different machines. 

News

What else is new

Register here for the ONDEWO newsletter. You will be up to date, and you will always be the first to know about new products and solutions as well as other news.