Styria Media’s AI transformation

By Scott M. Scher, Columbia University

Styria started its data team in 2014 and is Croatia’s biggest media company. Styria has split their AI efforts into two parts. AI for classifieds using computer vision and AI for news using natural language processing.

Computer vision analyzes image similarity and image classification. With this they have created willhaben, fashion cam. This allows a user to search by photo or existing advertisements to find similar products. The algorithm is able to recognize color, texture, and even brand recognition. Based on this a customer will be able to find specific products from a simple photo. Marko Velic, head of data for Styria, gave the example of being able to find specific Nike shows from a simple photo. Their product was able to search not only by type of product or color but also by brand. Styria is expanded this product from fashion into the antique and furniture markets as well. Willhaben had significantly better results than a typical Google search, he said.

On the seller side, they can determine make, model, and even the name of a product from an image. Their product has a 90 percent rate of correct answers and only a 10 percent response of “I don’t know.” Using their natural language processing, Styria are able to create recommendations based on these results.

Velic gave the audience a look into how an organization can start with AI. His first advice was to fail fast, fail cheap. To use AI to empower existing products rather than try and create something new. The benefit of AI to the organization is optimized processes, market targeting, and content personalization.