GeoAI for Knowledge Graph Construction: Identifying Causality Between Cascading Events to Support Environmental Resilience Research

Yuanyuan Tian,Wenwen Li

DOI: https://doi.org/10.48550/arXiv.2211.06011

2022-11-11

Abstract:Knowledge graph technology is considered a powerful and semantically enabled solution to link entities, allowing users to derive new knowledge by reasoning data according to various types of reasoning rules. However, in building such a knowledge graph, events modeling, such as that of disasters, is often limited to single, isolated events. The linkages among cascading events are often missing in existing knowledge graphs. This paper introduces our GeoAI (Geospatial Artificial Intelligence) solutions to identify causality among events, in particular, disaster events, based on a set of spatially and temporally-enabled semantic rules. Through a use case of causal disaster events modeling, we demonstrated how these defined rules, including theme-based identification of correlated events, spatiotemporal co-occurrence constraint, and text mining of event metadata, enable the automatic extraction of causal relationships between different events. Our solution enriches the event knowledge base and allows for the exploration of linked cascading events in large knowledge graphs, therefore empowering knowledge query and discovery.

Artificial Intelligence

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **When modeling disaster events, existing knowledge graphs lack the identification and linking of causal relationships between cascading events (cascading events)**. Specifically, current knowledge graphs usually only focus on single, isolated events, ignoring the mutual influence and causal connection between multiple events. This limitation restricts our understanding of complex disaster event chains, and further affects the effectiveness of disaster response and preparedness.

### Main problems of the paper
1. **Limitations of existing knowledge graphs**:
- When modeling disaster events, existing knowledge graphs are often limited to single events and lack the modeling of causal relationships between multiple events.
- There is a lack of systematic analysis and linking of cascading events (such as one event triggering another event).

2. **Methods that need improvement**:
- A new method that can automatically identify and link cascading events needs to be developed to improve the understanding of disaster event chains.
- By introducing Geo - spatial Artificial Intelligence (GeoAI) technology and combining spatio - temporal semantic rules to identify and model these causal relationships.

### Solutions
To overcome the above problems, the paper proposes a GeoAI - based solution, using a set of spatio - temporal semantic rules to identify the causal relationships between disaster events. Specific steps include:

1. **Topic - related event identification**:
- Expand the query scope through topic classification (such as "flood topic", "hurricane topic") to ensure that all relevant event types can be retrieved.

2. **Disaster matching under spatio - temporal co - occurrence constraints**:
- Use keyword matching and spatio - temporal co - occurrence constraints to associate specific disasters with related events. For example, use a time buffer (such as seven days before and after) to capture the impact range of events.

3. **Metadata text mining**:
- By mining causal relationship words (such as "cause", "trigger", etc.) in event metadata, construct an event - causal network to further determine the causal connections between events.

### Application cases
The paper shows how to use these rules to identify and link cascading events through a specific case - the flood and heavy rainfall events during Hurricane Sandy. Through this method, researchers can better understand the disaster event chain and provide more effective support for disaster response.

### Conclusions and future work
This research proposes a new GeoAI - based semantic analysis method, which can discover the inter - relationships between various severe weather events and their impacts on the environment and society. Future work will further optimize spatio - temporal co - occurrence constraints, introduce deep - learning technology for causal relationship prediction, and expand the method framework to support the construction of larger - scale knowledge graphs.

### Summary
This paper aims to solve the problem that existing knowledge graphs lack the identification of causal relationships of cascading events when modeling disaster events by introducing GeoAI technology and spatio - temporal semantic rules, thereby improving the understanding of complex disaster event chains and enhancing the disaster - resistance ability of cities and ecosystems.