Methods
“The system mimics human expert reasoning, making qualitative predictions based on comparisons between entities.”
“The system mimics human expert reasoning, making qualitative predictions based on comparisons between entities.”
Model Example : Slides In Soil
Model – Property – Property Value – Frequency
- slides in soil – has surficial material – morainal material (till) – always
- slides in soil – has surficial material – bedrock – sometimes
- slides in soil – has surficial material – colluvium – always
- slides in soil – has geomorphic process – erosional Process – always
- slides in soil – has geomorphic process – mass movement – always
- slides in soil – has slope – plain- rarely
- slides in soil – has slope – gentle – rarely
- slides in soil – has slope – moderate – usually
Minerva Machine Intelligence
The Minerva Intelligence reasoning approach is an expert-based qualitative and probabilistic matching system. The AI system mimics human expert reasoning, making qualitative predictions based on comparisons between models and instances. Models are conceptual abstractions of a given phenomenon (e.g. a landslide) described by a semantic network of properties, property values, and frequency terms (e.g. soil slide – has slope – steep – always). Instances are also described by semantic networks and are real-world representations described by true-or-false statements (e.g. slope unit – has slope – steep – true). Matching between instances and models is possible because the data are structured in the form of semantic networks, which are composed of hierarchical standardized taxonomies. The taxonomies we used are INSPIRE-compliant wherever possible. For an extensive review of the Minerva Intelligence matching methodology, see (Sharma et al., 2010, Smyth and Poole, 2004). We deliver, and portray on our output maps, a percentile rank between models and instances, that can be mapped as the relative spatial likelihood of landslide occurrence (i.e. susceptibility). A higher percentile rank, between an individual mapping unit and a landslide model, signals a higher susceptibility to that type of landslide over other landslides.
Minerva Machine Intelligence
The Minerva Intelligence reasoning approach is an expert-based qualitative and probabilistic matching system. The AI system mimics human expert reasoning, making qualitative predictions based on comparisons between models and instances. Models are conceptual abstractions of a given phenomenon (e.g. a landslide) described by a semantic network of properties, property values, and frequency terms (e.g. soil slide – has slope – steep – always) . Instances are also described by semantic networks and are real-world representations described by true-or-false statements (e.g. slope unit – has slope – steep – true). Matching between instances and models is possible because the data are structured in the form of semantic networks, which are composed of hierarchical standardized taxonomies. The taxonomies we used are INSPIRE-compliant wherever possible.For an extensive review of the Minerva Intelligence matching methodology, see (Sharma et al., 2010, Smyth and Poole, 2004) . We deliver, and portray on our output maps, a percentile rank between models and instances, that can be mapped as the relative spatial likelihood of landslide occurrence (i.e. susceptibility). A higher percentile rank, between an individual mapping unit and a landslide model, signals a higher susceptibility to that type of landslide over other landslides.
Model Example : Slides In Soil
Model – Property – Property Value – Frequency
- slides in soil – has surficial material – morainal material (till) – always
- slides in soil – has surficial material – bedrock – sometimes
- slides in soil – has surficial material – colluvium – always
- slides in soil – has geomorphic process – erosional Process – always
- slides in soil – has geomorphic process – mass movement – always
- slides in soil – has slope – plain- rarely
- slides in soil – has slope – gentle – rarely
- slides in soil – has slope – moderate – usually