Humans tend to think in analogs. For example, I did something during situation A...should I do that same thing during situation B? This depends on how similar the situations might be, either perceptually or in actuality. So goes the theory of analogies and, computationally, case based reasoning.
Wouldn't it be great if we could apply this to evaluations of how similar two or more places are? For example, from the perspective of a user, how similar are Washington DC and Dallas? It depends on what attributes a user cares about...maybe it's the restaurant scene, or how big the hair is, or how susceptible to infectious disease a particular area is?
That's what I'd like this project to be.
- Develop disease likelihood use cases
- Integrate existing ontologies (VectorBase, EnvO, etc.)
- Investigate suitable backend (Neo4j, OrientDB, etc)
- Google Knowledge Graph search results?
- Find data (DBPedia, GKG, etc.)
- Scientific justification - similarity metrics related to geospatial coverage, attributes, links to other things...
- topic modeling using gensim or similar
- write prototype in Python