System for automated search and download of contract terms

Product description

The system automatically downloads the terms and conditions of the website, including all linked texts. The result is a plain text format without tags that is easy to read and understand. The first collection is created based on keywords and is used to automatically create training data using machine learning methods. In the data thus obtained, keywords are masked for training the neural network and other regularization techniques are used so that the network does not learn simple rules and instead "understands" the text. This method is able to remove 71% of the errors of the keyword-based approach.