GENTIO Project Logo

GENTIO – Deep Learning Project

Generative Networks for Text and Impact Optimization

Project Start: 01 January 2020
Consortium: webLyzard technology, MODUL Technology, Vienna University of Economics and Business, Ketchum Publico and Observer
Funding: Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology (BMK), ICT of the Future Program

Advances in Deep Learning and Knowledge Extraction

Recent years have shown major advances in the use of deep learning for the automated extraction of factual, affective and contextual knowledge from digital content streams. GENTIO builds on these advances to change the way we produce, enrich and analyse digital content. The project will develop a flexible Multi-Task Learning (MTL) approach based on Generative Learning Networks to unify the understanding of text at three fundamental levels: structure, content and context. Thereby it aims to boost the context processing capabilities of Natural Language Processing (NLP) frameworks, reduce the high cost of developing training data, and support the cost-effective development of intelligent semantic systems. By offering interactive visualizations to explore the extracted features, the project will also put special emphasis on increasing the transparency of the underlying computational processes, which is a typical shortcoming of Artificial Intelligence-based systems. In this regard, the GENTIO consortium will closely work together with the CIMPLE Explainable AI project.

Use Cases: Marketing and News Production

Supported by multilingual and highly scalable knowledge graph technology, the envisioned approach will be applicable across numerous domains and regions. To demonstrate its versatility, two distinct domains have been chosen. The first use case targets the marketing domain. It will experiment with new methods for communication experts to maximize the impact of data-driven publishing. The second use case targets the news media sector, automatically correcting and classifying noisy output from Optical Character Recognition (OCR) systems – using topics extracted from the public debate on other micro-blogging sites to obtain the required context information.

Cross-Domain Exploitation Potential

The two use cases allow GENTIO to investigate the production as well as the analysis of digital content, driven by leading use case partners in their respective fields – Ketchum Publico as the Austrian representative of a global communications consultancy and OBSERVER as an established Austrian media intelligence SME with a history of more than 100 years. As part of the exploitation planning in the second half of the project, GENTIO will clearly define the potential of using its deep learning approach in a variety of other domains including broadcasting (semantic search for video retrieval), retailing and consumer brands (reputation management), telecommunications (helpdesk and support), consulting and auditing (legal text annotation and evaluation), and mobility (crowd-based feedback systems for autonomous driving applications).