We’re constantly improving the Re:infer platform, extending its capabilities and making it easier to use across the enterprise. Take a look at some of our biggest launches and improvements from the last quarter…
New Entity Model
Earlier in the quarter, we launched a completely new model for all trainable entities in the Re:infer platform. Our new entity model is built on a transformer-based architecture - a type of deep learning that learns to only pay attention to the most relevant parts of an input.
Users have been enjoying greater accuracy, faster training and better validation statistics, helping models enter production and start generating value much quicker. The foundation laid by the new model also opens up many exciting opportunities for deeper extractions in the future. Watch this space!
Specific Customer Domains
We’ve retooled how domains are managed in the platform. There haven’t been any changes to how users interact with Re:infer, but now when they log in users are automatically directed to a domain that is entirely specific to their company.
This paves the way for important new features and functionalities in the future, including the ability for users to create their own organisations or projects within their company’s domain.
See how the new login process works
Beamer Change Log
We’re very pleased to have launched a new, dedicated place for all of our platform updates as they happen. Our new product change log blog is hosted on Beamer, where new features, blog posts and platform announcements are posted as soon as they are launched.
Our Beamer change log is accessible publicly and in-app for our users, so you can keep up to date with all our new changes and capabilities in real time.
Work in Progress
Intelligent Extraction from Email
We’re hard at work on an exciting new beta feature for the Re:infer platform. We’re constantly experimenting with new ways to help customers create value from our natural language processing (NLP) models, with as little training as possible.
We’ve discovered that Re:infer’s NLP models can rapidly extract key information defined by the user from masses of unstructured email communications - all without any model training whatsoever. This opens up many exciting new use cases, enabling users to find the information they need instantly with Re:infer, without needing to sift through email messages manually.
The feature remains in beta as we continue to improve its accuracy and use-ability, but stay posted for more updates soon...
Pre-trained Label Concepts
We continue to find new ways of shortening time to value for our platform users. We have started to develop pre-trained labels that can be easily imported between data sets, and which can start making predictions and creating value immediately.
Labels improve in accuracy with training, but these pre-trained labels provide production-ready models that users can immediately pick up and start using straight ‘off-the-shelf’. Valuable insights can be generated in minutes, no upfront training required.
The current set of pre-trained labels is based on concepts that are common to email communications, such as ‘urgent’ or ‘complaint’, but we will continuously expand the number of options available.
Dynamic Thread View
We’re redesigning how messages are viewed in the Re:infer platform. Currently, messages are displayed individually, but soon users will be able to see the entire thread of messages that make up a conversation with the click of a button.
This will give users a whole new level of context on business communications, letting them quickly and easily analyse entire conversations with full visibility over every message.