Research

Artificial Intelligence (AI) is not only a fancy buzzword in our corporate name, but drives and influences us everyday. This includes but is not limited to finding and reading new exciting papers in the AI and machine learning field, compete with other data scientists on hackathons and Kaggle challenges or trying out new open source tools out there.

FH-Salzburg Logo

Besides that, with the ITS department of the Salzburg University of Applied Sciences (SUAS), we have a strong partner, with whom we are doing research projects together. Our partnership started in 2014 with the Article and Review Information Extraction (ARIE) project [2]. Since then, we have supervised several master thesis and Arnold Keller (CTO) and Martin Schnöll (AI & data scientist) started teaching Pattern Recognition as external lecturers. On a regular basis, we get together with our highly esteemed scientific colleagues exchanging ideas and discussing solutions for state-of-the-art problems and challenges.

Ongoing Research Projects

Digitalization and Data Analytics in the Energy Domain of Salzburg

Amongst others Fact AI is involved in the data analytics part of the DASPEOS project.

Time Series Analysis

Keywords: Recurrent Autoencoder (RAE), Long-Short-Term-Memory (LSTM)

Analyzing and comparing time series is a challenging task, mainly due to the variable length of such series. One, for example, want to compare the performance of different webshop users, including information like the order of single pages visited or how long a single user stayed on every single page. One state-of-the-art approach to encode such time series into representation vectors (codes) of same length, making them processable, is using Recurrent Autoencoders (RAE). Applying such a model on various datasets, including the analysis how such resulting codes can be compared semantically, is currently investigated and researched in a master thesis.

Publications

[1] Ferner C., Schnöll M., Keller A., Pomwenger W., and Wegenkittl S. (2016): Topic-Klassifizierung für automatisierte Produktbewertungen mittels Hidden Markov Modellen, Tagungsband 10. Forschungsforum der Österreichischen Fachhochschulen.
[2] Ferner C., Pomwenger W., Wegenkittl S., Schnöll M., Haaf V., Keller A. (2017): Information Extraction Engine for Sentiment-Topic Matching in Product Intelligence Applications. In: Haber P., Lampoltshammer T., Mayr M. (eds) Data Science – Analytics and Applications. Springer Vieweg, Wiesbaden.