Inductive Logic Programming (ILP) originated in the 1990s as a framework for combining symbolic learning with declarative knowledge representation. Today, mature ILP frameworks can learn complex, non-monotonic hypotheses, thus broadening the modeling capabilities and real-world applications of ILP. This work is primarily based on the FastLAS2 framework and aims to generate simple, interpretable hypotheses to clarify weather bulletins issued by OSMER FVG, the Regional Meteorological Observatory of Friuli Venezia-Giulia.
We present a pipeline that starts from simulated meteorological raw data and OSMER's bulletins (used as ground truth), extracting data as ASP facts and generating ILP examples. From these examples, an explanatory hypothesis is inferred via FastLAS2. This hypothesis is translated into natural language, explaining the weather forecast issued by human experts, particularly the rationale behind the experts' choices of specific symbols in the bulletin pictogram.
The proposed approach is general and can be applied to bulletins from other sources and regions.
Blogger's Review: The application of Inductive Logic Programming in weather forecasting showcases its powerful explanatory capabilities, distilling complex data into comprehensible conclusions, which is crucial for enhancing public understanding of meteorological information. The generalizability of this method also provides new insights for knowledge extraction in other domains.