Squirro Launches Dynamic Machine Learning Service, Including a Bespoke Training Data-Creation Module

November 14, 2018

Use cases include corporate financiers detecting exceptional or poor financial performance from earnings call reports

Zurich, 14 November 2018 – Augmented intelligence solutions provider Squirro has launched a new machine learning service that is one of the world’s first to include an Annotation Workbench that enables organizations to compose training data sets within the platform.

Squirro’s supervised machine learning classifier will learn to make predictions from labelled data, delivering deep insight from an organization’s unstructured data. Data scientists, analytics staff or even business users can now also add the training data that is so vital to machine learning, and can do so via the Annotation Workbench directly within the Squirro platform.

“Augmented intelligence and machine learning are becoming integral to many organizations, but machine learning is dependent on training data to be truly effective,” said Toni Birrer, Chief Technology Officer, Squirro. “Our new machine learning service is a major leap forward, offering the Squirro Annotation Bench that allows users to create training data, which can then be used to design, train and apply the machine learning model.”

Using the new Squirro machine learning service is straightforward. Users upload unstructured data and then build machine learning classification at document, paragraph or sentence level, classifying each with supplied tags. The algorithm will learn to identify and differentiate between the tags when analyzing sentences by learning the context of associated words to provide an even more accurate and insightful assessment.

One application of the new service is for corporate financiers to detect exceptional or poor financial performance from analyzing earnings call transcripts. The analysis of earning call transcripts, which can run to hundreds of pages long, is a lengthy and time-consuming process. By classifying sentences as ‘exceptional performance’ or ‘poor performance’, the same data can be used to deliver deeper insight much quicker and more effectively than if done manually.

“The power of machine learning to enhance the capacity and capability of humans is growing all the time, as new tools emerge, and businesses come up with new use cases,” adds Toni Birrer. “We wanted to provide a service that includes the creation of training data that can then be used to train the machine learning model. With our Annotation Workbench we have achieved that, opening up machine learning for a variety of use cases.”