Words are used to refer to objects in reality. One word can imply many references to different objects that are categorized as similar. For example, the word "city" can refer to Vienna, Alexandria, or Las Vegas; the word "near" can refer to a range of distances, e.g. "moon is near the earth" or "near St. Stephens cathedral". If a Geographic Information System (GIS) is queried with a sentence including "city" or "near", the challenge for an algorithm executed by the GIS is to decide which exemplar of the word "city" or which distance "near" refers to. To overcome this challenge, the hypothesis is that context selects references to objects in reality. A context algebra is presented, implemented, and used to represent the word "near", in order to evaluate the hypothesis. Context algebra makes use of the theory established by a context-enriched semiotic triangle. The semiotic triangle connects objects in reality to words via concepts in an agent. Context is represented with the mathematical structure lattice and is mapped to a contextualized concept. A contextualized concept is modeled with sets of objects observed from reality, where a typical object is determined. This typical object (prototype) is assumed to be the translation from a word to an object in reality. For example, the influencing context "capital of Austria" for the word "city" selects the prototypical instance Vienna. The initialized model for "near" makes it possible to determine references to different distances according to contexts, e.g. the size of the referenced object or activity. For example, "near the lake" references 1050 meters, "near by subway" references 380 meters, and "near by walking" references 550 meters. The results of the application for "near" provide evidence that the hypothesis is valid, and that it is able to select references which translate for example "near" into metric distances.