Determining Context

Here are some example rules contained in a Rulebase generated from the “Apollo 11” concept based on its relationships to other concepts:

1. The Associative Relationships are defined in the model. They can be typed custom relationships – as illustrated here – or “has related” relationships.
2. The Publisher applies the template and expands the “relationshiptypes” according the information contained in the model
3. When Classification Service (CS) is running the “Apollo 11” Rulebase, it will follow the resulting Link Rules to test if the document is about these related topics. If it is, extra score will be generated for “Apollo 11”.

The above diagram shows Link Rules, which are used to validate the context in which a word appears. For example, the taxonomy node for the British telecommunications company “Orange” would have an Associative Relationship in the taxonomy with the “Telecoms Sector” node. When the Publisher generates the Rulebase, it will create a Link Rule. When the “Orange” Rulebase is processed by CS, the Link Rule will make it test the document against the “Telecoms Sector” Rulebase as well. If the document mentions “Orange” and talks about telecoms, it is more likely to be about the company “Orange” than the fruit, color or county.

Link Rules also can be used to tag documents with the parent nodes. For example a document about “Apollo 11” would also be tagged with “Apollo Flights”, “Manned Space Flight” and “Space Flight.” This is optional and is configured in the Rulebase Template.

More details of on how the ontology and Rulebase Templates work together to classify documents can be found in the Semaphore documentation available via the customer portal.