November 23, 2023
Continuing the series of Invited Talks organised by SKEL | The AI Lab, the Institute will host Dr Eleni Tsalapati, Research Fellow in the Department of Informatics and Communications at NKUA, to deliver a talk titled Reasoning over ontological knowledge bases with standard and non-standard methods on Friday 1 December 2023, at 10.00.
The talk will take place physically at the Aegeo Meeting Room of the Institute of Informatics & Telecommunications, while participation will be also possible through zoom.
For online participation:
https://us02web.zoom.us/j/83124148533?pwd=ZUxIb1I3cmNQcVE2anIxY1VHcUZpQT09
Meeting ID: 831 2414 8533
Passcode: 591709
About the talk: In this presentation, we will initially focus on scalable techniques for query answering over inconsistent DL-Lite knowledge bases. These techniques are based on the IAR and ICAR semantics proposed in the literature. Furthermore, an algorithm that computes upper approximations is presented for the case of more expressive DLs, where the problem is known to be intractable.
A further development pertains to the fact that reasoning over expressive logic languages is a time-consuming process while expressing knowledge formally is an arduous task for domain experts. To overcome these bottlenecks, we explore how we can utilize transformer-based models. One way that the current state of the art measures the reasoning ability of transformer-based models is by evaluating accuracy in downstream tasks like logical question answering or proof generation over synthetic contexts expressed in natural language. However, most of the contexts used are in practice very simple; in most cases, they are generated from short first-order logic sentences with only a few logical operators and quantifiers. Here, we present how well a transformer-based model will perform reasoning over expressive contexts (generated from the description logic language ALCQ). Finally, a use-case scenario of semantic technologies is presented from the fuel cell system diagnostics domain.
Speaker Bio: Dr Eleni Tsalapati is a Marie Curie Research Fellow in the Department of Informatics and Communications at NKUA. She brings over 12 years of experience in knowledge representation and reasoning. . Currently, she explores ways to employ the capabilities of LLMs to approach efficient and effective question answering over expressive knowledge graphs. Throughout her research career, she has actively been involved in 10 projects (ESA, European, EPSRC, HFRI). Since she joined Manolis Koubarakis’ team, in 2020, she has also been involved (as a researcher/project manager) in various projects (ESA, GeoQA, AI4Copernicus) pertinent to knowledge representation and retrieval of earth observation/geospatial data.