PrediCX is a super charged text classification machine, able to classify text pieces based on ‘human-in-the-loop’ input. PrediCX allows our customers to classify millions of pieces of text like warranty claims, customer ticket support, chatbot interactions, product reviews, or customer complaints, in order to automate customer feedback inside, customer support prioritization or warranty handling.
How does it work?
Glad you asked! Traditional machine learning approaches to labelling require an entire large, static dataset to be labelled manually, PrediCX achieves the same classification performance but requires far less labelling. This is achieved by requesting a human to manually label specific records of the dataset. This is a continuous process, with a human input which maximizes labelling performance while guaranteeing minimal manual labelling effort.
In traditional labelling approaches using a static dataset,the labelling performance tends to degrade over time as the data changes - new concepts are introduced, and old concepts are described using varying nomenclature and phrasing. However, continuous input from a human allows Predicx to overcome these problems.