Q3 2021 Platform Update - What’s new in Re:infer
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…
Mixed Language Models
Re:infer machine learning models are now able to support sources and datasets containing multiple European languages - no translation required. Users can now train models on data that has a mix of European languages, even within the same conversation. We’ll consistently be adding to the number of languages we support.
This greatly expands the range and diversity of use cases Re:infer can serve. The conversations happening at large enterprises are multi-channel and multi-lingual. Re:infer users will now be able to analyse and understand conversational data from across the world.
To make the model training process quicker and easier in Re:infer, we have developed Model Rating. This is a simple but powerful feature that analyses the overall health and performance of machine learning models created on the platform.
Subject matter experts (SMEs) devote much time and effort to the essential task of auditing and reviewing their machine learning models. Model Rating streamlines this process by analysing models based on their accuracy, balance and performance. It then reports back a holistic score that gives SMEs an accurate picture of the model’s health, as well as the actions needed to improve it. This saves precious time in the training phase and helps models start generating value faster.
We’ve made the entities that users create in Re:infer even smarter. The training process has been streamlined as entities can now inherit the structure of other entities, helping to formalise an entity structure that is useful, actionable and specific for the business. Not only does this improve model accuracy, it reduces training time and helps models reach deployment faster.
Advanced Prediction Filters
We’ve augmented the platform’s analytical capabilities to allow users to apply multiple label filters at once, exclude certain labels with a filter, search with label filters, and more.
This feature was in high demand among our customers, and now enables them to perform deeper analysis of their communications data. Users can query the data on any collection of meta-data and labels, enabling them to - for example - discover the most common requests made by customers who’ve had a bad experience using the company website.
Compare UI refresh
Our Compare feature enables users to analyse and compare groups of messages - ideal for testing the impact of a new marketing initiative or change programme across different regions, customer groups and time periods.
Following our redesign of the Compare page, this process will be even quicker and easier. We’ve added to the feature’s functionality with an interface that’s simpler and more intuitive.
We’re proud to have launched a dedicated web portal for developers. The Re:infer Developer Portal contains useful information for the development community, including API documentation and practical integration guides for the platform.
The site also provides a space for our own experts, architects and engineers to share their tips, insights and best practices on our Developer Blog.
Work in progress
The team is hard at work developing pre-trained concepts that will further shorten the training time needed to develop models on the platform. Stay tuned for more changes and updates soon.