This is a simple implementation of the propagation-based algorithm for question answering in sementic networks.
We'll take this example to illustrate all the process from the creation of the graph to the question answering.
The graph is constructed by giving a serie of statements.
For our example, the statements would be as follows :
- Asian is_a Human
- European is_a Human
- Chinese is_a Asian
- Japanese is_a Asian
- French is_a European
- French eat snail
- Japanese eat rice
- Chinese eat Grasshopper
- Grasshopper is_a Insect
It is very important to keep the exact same spelling for relations and actors. Also, is_a needs to be in this exact spelling.
This will create the following graph structure :
{'is_a': [('Asian', 'Human'), ('European', 'Human'), ('Chinese', 'Asian'), ('Japanese', 'Asian'), ('French', 'European'), ('Grasshopper', 'Insect')], 'eat': [('French', 'snail'), ('Japanese', 'rice'), ('Chinese', 'Grasshopper')]}In order to answer the question "Which Humans eat Insects" we write : Human eat Insect. The answer is given to us as: < Chinese eat Grasshopper > since chinese are Human, and Grasshoppers are insects.