Financial tasks & Deep learning the perfect fit

Why Deep Learning?

/ provide better generalization than rule based systems

/ efficient maintenance and updating

/ the perfect technology for automating repetitive processes

SMACC's AI technology applied

/ more than 300,000 documents

/ more than 25,000 different layouts


SMACC uses Artificial Intelligence based on deep learning technology to automate financial tasks. Here's how.

SMACC's technology applied


From training data to automated use of model

In the first step we are extracting the raw text from a PDF document or in case of paper documents we work with an optical character recognition (OCR) system. The OCR recognizes text from scanned invoice documents.

A human labels the input data, e.g. in the case of invoices, certain text snippets in the input data are labeled as “IBAN”, “BIC” or „TAX-RATE“.  The algorithm is trained on the data to find the  relation between the input features and the output labels. Multiple hierarchical layers (deep learning) improve representational power of the model.

Once training is complete the built model is applied to new data, i.e. even unseen documents will be labelled automatically.

The Evolution of Artificial Intelligence (AI)

AI is typically defined as the ability of a machine to perform cognitive functions we associate with human minds, such as reasoning, learning, interacting with the environment, and problem solving. Most recent advances in AI have been achieved by applying machine learning techniques to very large data sets. Machine learning is considered as the main technology used in order to achieve AI. Deep Learning is one of the many machine learning approaches. It enables software to train itself to perform tasks, such as speech or image recognition, by exposing multilayered neural networks to vast amounts of data. Deep Learning has been the most successful technology in industry in the last few years. In the following, we will deep dive into how the technologies work.


Machine Learning

Machine-learning algorithms detect patterns and learn how to make predictions and recommendations by processing data and experiences, rather than by receiving explicit programming instruction. The algorithms also adapt in response to new data and experiences to improve efficacy over time.
Supervised Learning
Supervised learning algorithms use training data labeled by humans to learn the relationship of given inputs to a given output. The trained models are able to infer the correct output for new and even unseen input data. They are also capable of making predictions and are therefor also often used in predictive modelling and predictive analytics.
Unsupervised Learning

Algorithms explore input data without being given an explicit output variable. Unsupervised learning algorithms find patterns and structures in unlabeled data.


Deep Learning with Neural Networks (DNN)

DNN with multiple hidden layers are used to understand big amounts of data. Deep learning is a type of machine learning that can process a wider range of data resources, requires less data preprocessing by humans, and can often produce more accurate results than traditional machine-learning approaches (although it requires a larger amount of data to do so). In deep learning,
interconnected layers of software-based calculators known as “neurons” form a neural network. The network can ingest vast amounts of input data and process them through multiple layers that learn increasingly complex features of the data at each layer. The network can learn the most important features of the training data and is able to infer the correct labels of new data. For example, once it learns what an object looks like, it can recognize the object in a new and unseen image.
Recurrent Neural Network (RNN)

A (multilayered) neural network that can store information in context nodes, allowing it to learn data sequences and output a number or another sequence.
Used for the SMACC data extractor.

Convolutional Neural Network (CNN)

A multilayered neural network with a special architecture designed to extract increasingly complex features of the data at each layer to determine the output. Mainly used for image recognition.


„Artificial Neural Networks enable the automation of complex human tasks. Deep Learning Technology learns from historical data and can apply these learnings in a reliable manner. This technology is ideal for financial services.“

Dr. Ulrich Erxleben, Co-Founder SMACC

Get in touch

Please note our privacy policy.