Segmentation and Aggregation in Text Classification

In the context of AI systems for industry, analysis of text becomes an important part. One application is the classification of documents. While multiple classification algorithms are known, in application those suffer from the different length of the documents. One way to handle those is the segmentation setup. As a result, the classification algorithm gives a probability for each segmentation and every class. To classify the whole document, an aggregation method is needed. We investigate different approaches of aggregating and apply those methods to text documents for public administration.