Every year the Extended Semantic Web Conference (ESWC) is one of the leading destinations for many researchers in the Semantic Web community. The ESWC event also offers a PhD symposium: a forum for PhD students in the area to present their work and an opportunity to receive constructive criticism from mentors and peers. This year’s edition of the symposium was organised by Dr Ilaria Tiddi (Vrije Universiteit Amsterdam, The Netherlands) and Claudia D’Amato (Università di Bari, Italy). Agnese Chiatti, PhD Research Student at KMi, presented a technical overview of her PhD work titled "Towards Visually Intelligent Agents (VIA): a Hybrid Approach".
Agnese’s PhD work proposes a framework of epistemic requirements for developing Visually Intelligent Agents, i.e., to support the ability of service robots to make sense of the environment through their vision system, reasoning components and background knowledge. Intermediate results from this work show that a hybrid reasoning architecture that introduces awareness of objects’ typical size can significantly augment object recognition solutions purely based on Deep Learning. Agnese also presented a novel First Order Logic framework to represent the typical Qualitative Spatial Relations between objects which can be fully implemented through Geographic Information Systems (GIS) libraries. This spatial component will be used further to augment the proposed hybrid architecture on visual reasoning tasks.
Agnese’s virtual presentation of her work received great interest from the symposium audience, leading to a very engaging Q&A session that spanned for 40 minutes after the presentation. The ESWC PhD symposium is an excellent opportunity for students to receive valuable feedback and practice defending their work in a friendly and engaging environment. KMi has a long history of students and scholars who have actively participated in this event, and many present and future KMi students may benefit from joining this initiative.