This project implements algorithms for determining the relatedness between two concepts and for annotating text based on concept relatedness.

Given a knowledge base or ontology as input, the relatedness between two concepts is computed using:

• a definition-based measure: based on the concept and connected concepts definitions (human-readable text describing their meaning)
• a structure-based measure: using the ontology or knowledge base structure (how concepts are related in the knowledge base)
• a hybrid measure combining concept definitions and knowledge base structure

Words in text are annotated with concepts from the ontology or knowledge base as follows. First, candidate concepts are identified for words in a given context window. Second, the relatedness between candidate concepts in the context window is computed. Third, the candidate concept with the best relatedness score is selected to annotate the word.

The code is organized in four main packages:

1. The knowledgebase package contains modules for representing the knowledge base as a NetworkX graph of concepts and relations between concepts, as well as a module for representing concept definitions as a Bag of Words (BOW).
2. The text package is useful for text pre-processing.
3. The relatedness package modules contain implementations of the definition-based and structure-based algorithms.
4. The annotation package contains modules with the implementation of the text annotation algorithm.

See util.writeconfig for the parameters that should be set for the relatedness and text annotation modules.

The relatedness and annotation packages contain example code for determining the relatedness between two concepts and for annotating text respectively. See relatedness.main_relatedness and annotation.main_annotation for more details.