GENTIO conducts experiments using state-of-the-art machine learning, closely aligning its datasets and experiments to the use cases driven by the partners Ketchum Publico and Observer. After ensuring that sufficient high quality data was available, the project evaluated different configurations of generative network architectures for multiple tasks, and extended its capabilities to visualize the inner workings of our data collection and deep learning pipelines.
WP1: Project Management entails strategic and day-to-day management and coordination activities, as well as ongoing communication with the funding agency as the funding agency and other professional stakeholders.
WP2: Knowledge Integration and Modelling. To provide the required datasets, this work package improves the accuracy of data annotation and analysis methods through (NLP, NER) contributing to the underlying knowledge graph.
WP3: Generative Networks has implemented a set of generative language models based on state-of-the-art architectures like Transformers, and comparatively evaluated these models to gain insights into the best network architecture for specific use case tasks.
WP4: Feature Visualization and Dashboard Integration has implemented the first sets of visualizations of the workings of the deep learning networks of WP3 and began integrating them into GENTIO’s visual analytics dashboard.
WP5: Data-Driven Publishing has provided insights into user requirements for tools that can optimize the impact of user-generated content, including UI/UX specifications and user evaluations to assess the current interfaces of the dashboard and Storypact Editor.
WP6: Text Correction and Classification has implemented a first version of OCR text correction tools including REST API and visual frontend to enable an evaluation of results and identification of remaining errors.
WP7: Dissemination and Exploitation Planning increases the impact and visibility of the project in the research community and vis-à-vis professional stakeholders, and plans the exploitation of results beyond the project’s lifetime.