An illustrated approach to Soft Textual Cartography

Area of Science:

• Geographic Information Science
• Computational Linguistics
• Data Mining

Background:

• Geographic information retrieval often struggles to integrate diverse textual data with spatial configurations.
• Understanding regional characteristics requires analyzing both location and descriptive content.

Purpose of the Study:

• To develop and demonstrate Soft Textual Cartography (STC), a novel approach for clustering geographical regions.
• To integrate spatial relationships and textual descriptions for enhanced pattern discovery.

Main Methods:

• Constructing a weighted network representing geographical layout.
• Utilizing topic modeling to extract thematic dissimilarity from textual data.
• Applying a soft k-means algorithm with expectation maximization and label propagation for soft clustering.

Main Results:

• Successfully applied STC to Wikipedia data of Swiss municipalities.
• Revealed significant textual autocorrelation and associations with official classifications.
• Demonstrated the potential of integrating spatial and textual data for topic interpretation.

Conclusions:

• Soft Textual Cartography offers a synergistic approach to geographical information retrieval and topic interpretation.
• The method effectively reveals spatial patterns by merging textual content with geographical context.
• The approach allows for the incorporation of expert knowledge into the clustering process.