semantic-network 0.2.0

Semantic network implementation

semantic_network Package

SemanticNetwork

The class is designed for encoding graph structures using universal triplets:

entity_1 relation entity_2
entity_2 relation entity_3

• Triplet parts are separated by spaces
• First and third parts are objects
• Second part is a directed relationship between objects

1. Creating and Extending the Graph:

from semantic_network import SemanticNetwork

script = """
    Circle hasPart Circle.radius
    Circle hasPart Circle.center
"""

sn = SemanticNetwork.from_script(script)

script_2 = """
    c1 fromProto Circle
    c1 hasPart c1.radius
    c2 fromProto Circle
    c2 hasPart c2.radius
    c2 hasPart c2.center
    c2.radius fromProto Circle.radius
"""

sn.append_script(script_2)

This creates a graph with objects and relationships:

print(sn)
----------------------------------------
|- c1
|- Circle
|- c1 fromProto Circle
|- c1.radius
|- c1 hasPart c1.radius
|- c2
|- c2 fromProto Circle
|- c2.radius
|- c2 hasPart c2.radius
|- c2.center
|- c2 hasPart c2.center
|- Circle.radius
|- c2.radius fromProto Circle.radius
----------------------------------------

2. Graph Search

Search queries are specified using a special query format. Elements starting with * are treated as variables, while other elements are treated as constants.

q = """
    *o1 hasPart *o2
    *o1 fromProto Circle
"""

for objs, rels in sn.query(q):
    print(objs, rels)

Returns dictionaries with found graph fragments:

{'o1': 'c2', 'Circle': 'Circle', 'o2': 'c2.radius'} {'fromProto': 'fromProto', 'hasPart': 'hasPart'}
{'o1': 'c2', 'Circle': 'Circle', 'o2': 'c2.center'} {'fromProto': 'fromProto', 'hasPart': 'hasPart'}
{'o1': 'c1', 'Circle': 'Circle', 'o2': 'c1.radius'} {'fromProto': 'fromProto', 'hasPart': 'hasPart'}

During search, the system also verifies that the query graph is connected and acyclic.

3. Utility Methods

# Extend the graph
sn.append_script(new_script)

# Print graph description (number of unique objects and relationships)
sn.describe()

# Create a copy
sn.copy()

# Serialization to dictionary and creation from dictionary
sn.to_dict()
sn = SemanticNetwork.from_dict(dict_data)

# Write to file with additional custom variables dictionary, and read from file
sm.dump(path, variables)
sn = SemanticNetwork.load(path)