Artificial Intelligence, AI for short, has been on everyone’s lips not just since ChatGPT. Machine learning holds huge potential for the IT industry. ITscope also relies on machine learning methods.
The task: categorise 7 million products
On average, 100,000 new products from distributors and manufacturers across Europe find their way into the ITscope platform every month, and the trend is rising. To ensure that every customer finds exactly what they are looking for in the huge selection of over 7 million products, all items are uniformly categorised and sorted. Since November 2022, ITscope’s content managers have been supported by a machine learning model.
The algorithm: learning from data sets
Machine learning, abbreviated ML, is a sub-discipline of AI and allows computers to learn from data sets and mimic human decisions. On the ITscope platform, an algorithm has been in use since November that processes product data autonomously. Previously, content managers categorised the products using stored rules and then checked them manually. The new algorithm was trained with the existing data and largely autonomously assigns the products or suggests suitable categories.
Over 7 million products from various manufacturers and distributors are categorised in ITscope.
The reality: ML models need people
So artificial intelligence is now replacing employees at ITscope? It’s not quite that simple: Although the ML model can correctly assign a high percentage of products, there is still a lot of product data that cannot be clearly categorised. At this point, our content managers intervene again and assess the data with their IT expertise. The ML model learns from these decisions and thus becomes more and more accurate. In this way, artificial intelligence ensures that the workload in content management remains constant despite the ever-growing number and variety of products.
The trust: Creating transparency
To ensure that the algorithm really makes the content managers’ work easier, ITscope got support from the Hochschule Ruhr-West: Jonas Deterding dealt with the explainability of the categorisation model through Explainable AI techniques for his bachelor thesis. “Acceptance and trust form two important aspects that can both support and prevent the use of an ML model,” explains the Bachelor graduate. “Through explainability, the user gains an improved understanding of the model predictions, so that he does not have to trust them blindly.”
The result: Efficient processes
Product Manager Jan Crommelinck is convinced by the innovations of the ITscope platform. “The algorithm takes the easier categorisation tasks off our colleagues, so they can focus on the more complex cases. ML is also excellent for checking existing categorisations and correcting errors in a targeted manner, if necessary,” he says and is also pleased about the cooperation with the university: “Jonas ensured that AI decisions become transparent with his bachelor thesis, which has accelerated our workflows.” This is also reflected in figures: with ML support, each content manager is almost twice as fast compared to the same period last year without ML support.