Semantic indexer

This is the core of our innovation and what makes our technology very reliable: on average, 1000 semantic signals for a 100-word message.

Each language is modeled dynamically, using Behavioral Patterns: they represent the probabilities of assembly of each word with all others, and also with the knowledge already present (the semantic dimension). 

This is how our engine finds the meanings of each word, and assembles these meanings together. This is how it learns continuously, by building itself new Behavioral Patterns. 

This is how it adapts not only to each language, but to YOUR particular context..

Semantic manufacturing of business data

Each business area and sometimes each company has its own vocabulary. And these particular words have precise meanings, known by your agents but not always by your customers.

After Semantic Indexer, that produces all the signals present in messages, Categorization transforms this overly rich mass into precise, structured data that can be used by your agents and your processes.

On average, 12 business data for a 100-word message.

Machine learning

The particularity of the OWI technology is our double machine learning: linguistic ML during the Indexation (instant learning) and business ML, periodically or on demand (Gradient descent, via Pytorch). 

It is this double technology that makes our machine learning extremely fast, and brings an outstanding reliability. It also allows the automatic discovery of new signals. 

Our business ML is supervised and fully explainable.


One of the challenges of AI is learning data.

Our solution allows us to directly capture the choices of customers and agents and transform them into “unitary signals”, automatically exploitable by our Machine Learning, without human intervention. 

Thanks to this, the solution can be deployed very quickly, and is continuously improved, always under your supervision. 

Learning transfers and data protection

Without large volumes of data, there is no machine learning. But how do you reconcile this with your need to protect your customers’ data?

Each OWI Client has a dedicated environment where data is isolated and protected. 

Each learning is done within an environment, only with the data that is present. 

Learning transfers allow you to easily integrate into your environment parts of models from learning done in other environments.

Thanks to learning transfers, we offer you packages developed on hundreds of millions of messages. 

The best third-party technologies

Focused on our expertise, we have partnered with the best technologies :

– Voice (Microsoft, Google, Nuance… etc)
– Document and image analysis (Abbyy, Kofax… etc)
– Contact management (WL Contact, GenesysEnghouse… etc)
– CRM (Salesforce, Microsoft… etc)

We also offer a wide range of connectors so that you get the maximum value from our solution, whenever you need it. And so that with each new use case you capitalize a little more on a richer and more efficient learning experience.