Midas: a Python Framework for Automated Generating and Training of Neural Network Models
Pedersen, Nikolaj Krebs, Anders Widtfeldt Meged, Rune Johansen · IT University Of Copenhagen (IT University of Copenhagen) · 2022
As social interactions increasingly take place in digital environments, a vast and increasing volume of digital traces in the form of unstructured textual data is produced. Currently, the machine learning techniques to analyze such data require deep knowledge of machine learning, usually confined to highly specialized data scientists and statisticians. This high competence threshold excludes a large group of businesses and data analysts from leveraging machine learning and creating value from unstructured digital trace data. This paper shows how the Midas framework can help data analysts analyze unstructured text data using neural networks. It outlines the framework's main features and provides example guides on how to implement the Midas framework on specific datasets. We hope this framework will make it easier and more accessible for data analysts working with unstructured text data to leverage neural networks.