Analyzing huge amounts of data, making predictions, and providing personalized recommendations: Machine learning algorithms are able to do in a very short time what no human can do. This special ability is used in various areas of life to handle tasks and make decisions even more efficiently. But why exactly a machine arrives at certain results, we often cannot understand with our human judgment. HUBBLE, a research project, in which Spectos participated together with HTW Dresden and Mittweida University of Applied Sciences (HSMW), focused on this question.
Intelligent Text Classification with Machine Learning Algorithms
The goal of HUBBLE was to investigate the influence of humans on machine learning algorithms and to transfer this knowledge to real work contexts. For this purpose, the new algorithm approach HBBO, Human Behaviour-Based Optimization, was used. Based on this algorithm, the goal was to explore the behavior and the possibility of human influence on text classification and text categorization. The results of the project contribute to the development of AI models that can be better influenced. Such models can increase the accuracy of frequently performed trend analyses and make text-based processes more efficient.
One typical use case in real life are companies and service providers with large amounts of unstructured data, e.g., from free text of emails in customer support ticket systems. Thanks to classifications and categorizations via corresponding algorithms and partially and fully automated procedures, processing operations can be accelerated considerably. In the context of the research project, training data from the ticket system of a customer dialogue system from Spectos was used and the developed algorithm was examined with regard to text recognition and classification.
The two-year project (until February 2022) was funded by the Zentrales Innovationsprogramm Mittelstand (ZIM). In its role as an industry partner, Spectos has gained valuable experience in the areas of text recognition, AI, and machine learning algorithms. This knowledge can be applied in customer projects to further advance business process optimization based on partially and fully automated processes.