Abstract | ||
---|---|---|
People use verbal descriptions and graphical depictions to communicate spatial information, thus externalizing their spatial mental representations. In many situations, such as in emergency response, the ability to translate the content of verbal descriptions into a sketch map could greatly assist with the interpretation of the message. In this paper, we present an outline of a semi-automatic framework enabling seamless transition between verbal descriptions and graphical sketches of precinct-scale urban environments. The proposed framework relies on a three-step approach: NL parsing, with spatial named entity and spatial relation recognition in natural language text; the construction of a spatial Property Graph capturing the spatial relationships between pairs of entities; and the sketch drawing step where the identified entities are dynamically placed on a canvas in a manner that minimizes conflicts between the verbalized spatial relationships, thus providing a plausible representation of the described environment. The approach is manually demonstrated on a natural language description of a university campus, and the opportunities and challenges of the suggested framework are discussed. The paper concludes by highlighting the contributions of the framework and by providing insights for its actual implementation. |
Year | DOI | Venue |
---|---|---|
2013 | 10.1007/978-3-319-01790-7_17 | COSIT |
Keywords | Field | DocType |
dynamic sketching,nl place descriptions,spatial information extraction | Spatial analysis,Spatial relation,Computer science,Human–computer interaction,Artificial intelligence,Sketch,Named entity,Natural language,Parsing,Conceptual framework,Multimedia,Machine learning,Mental representation | Conference |
Citations | PageRank | References |
16 | 0.79 | 28 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Maria Vasardani | 1 | 117 | 15.67 |
Sabine Timpf | 2 | 299 | 40.18 |
Stephan Winter | 3 | 643 | 45.20 |
Martin Tomko | 4 | 181 | 21.96 |