Spatial Semantics

Printer-friendly version

Extracting Meaning from Lake Bathymetry

One of the goals of the Jefferson Project is to understand how Lake George's bathymetry might contribute to problems arising in the Lake. However, raw coordinates obtained from various remote sensing technologies (Lidar, sonar, etc) may not be very useful without some manipulation. In order to allow researchers as many tools as possible to make effective use of this data, including IBM's Watson, the data must be converted into a more readable form. Specifically in the case of Watson, the data must be converted into natural language text. Making this conversion allows researchers more access to the data, and allows the phrasing of questions in natural language.

In order to accomplish the conversion of numerical bathymetry data into natural language, the following steps are taken.

  1. Group numerical data (via unsupervised clustering), and identify points of interest (e.g high-points, low points, etc)
  2. Classify the objects which the clusters represent
  3. Determine how the classified objects relate to each other topologically
  4. Systematically convert the object descriptions and their relations into plain English
  5. Use the generated text as a knowledge base for Watson

Following the above blueprint, simple coordinates from the lake can be turned into simple statements. The table below is an example of how the resulting statements may be organized, and shows generally how the process can extract meaningful information hidden in the raw data.

example_table.jpg

When paired with outside knowledge of the effects of bathymetry on a Lake's systems, such natural language text could provide a quite valuable knowledge base for a question answering system such as IBM's Watson.