Yohei Kurata

Date of Award


Level of Access

Open-Access Dissertation

Degree Name

Doctor of Philosophy (PhD)


Spatial Information Science and Engineering


Max J. Egenhofer

Second Committee Member

M. Kate Beard-Tisdale

Third Committee Member

Werner Kuhn


People often sketch diagrams when they communicate successfully among each other. Such an intuitive collaboration would also be possible with computers if the machines understood the meanings of the sketches. Arrow symbols are a frequent ingredient of such sketched diagrams. Due to the arrows’ versatility, however, it remains a challenging problem to make computers distinguish the various semantic roles of arrow symbols. The solution to this problem is highly desirable for more effective and user-friendly pen-based systems. This thesis, therefore, develops an algorithm for deducing the semantic roles of arrow symbols, called the arrow semantic interpreter (ASI). The ASI emphasizes the structural patterns of arrow-containing diagrams, which have a strong influence on their semantics. Since the semantic roles of arrow symbols are assigned to individual arrow symbols and sometimes to the groups of arrow symbols, two types of the corresponding structures are introduced: the individual structure models the spatial arrangement of components around each arrow symbol and the inter-arrow structure captures the spatial arrangement of multiple arrow symbols. The semantic roles assigned to individual arrow symbols are classified into orientation, behavioral description, annotation, and association, and the formats of individual structures that correspond to these four classes are identified. The result enables the derivation of the possible semantic roles of individual arrow symbols from their individual structures. In addition, for the diagrams with multiple arrow symbols, the patterns of their inter-arrow structures are exploited to detect the groups of arrow symbols that jointly have certain semantic roles, as well as the nesting relations between the arrow symbols. The assessment shows that for 79% of sample arrow symbols the ASI successfully detects their correct semantic roles, even though the average number of the ASI’s interpretations is only 1.31 per arrow symbol. This result indicates that the structural information is highly useful for deriving the reliable interpretations of arrow symbols.