Data Science

Motivating Dependency Parsing

The DependencyParser assigns syntactic labels describing the relations between individual tokens
The parser also supports sentence boundary detection
This lets us iterate over base noun phrases (or chunks)
Noun chunks refer to a noun plus the words describing the noun
For example, the lavish green grass is a noun chunk

Sample Dependency Tree

>>> doc = nlp("Wow! Spacy is a great tool and I want to learn more.")

# Plot dependency tree
>>> displacy.serve(doc, style="dep")

DependencyTree

Relevant Functions and Attributes

Token.head: The syntactic parent of governor of this token
Token.children: A sequence of this token's immediate syntactic children
Token.ancestors: The rightmost token of this token's syntactic descendants
Token.lefts: The leftward immediate children of this token in the syntacic dependency parse
Token.rights: The rightward immediate children of this token in the syntactic dependency parse
Token.subtree: A sequence containing this token and all of its syntactic descendants

Navigating the Parse Tree

>>> doc = nlp("Wow! Spacy is a great tool and I want to learn more.")

# Find parent of token
>>> spacy_token = doc[2]
>>> is_token = doc[3]
>>> a_token = doc[4]

>>> spacy_token.head.text
'is'
>>> is_token.head.text
'is'
>>> a_token.head.text
'tool'

# Find all ancestors
>>> [i.text for i in spacy_token.ancestors]
['is']
>>> [i.text for i in is_token.ancestors]
[]
>>> [i.text for i in a_token.ancestors]
['tool', 'is']

# Find immediate children
>>> [i.text for i in spacy_token.children]
[]
>>> [i.text for i in is_token.children]
['Spacy', 'tool', 'and', 'want']
>>> [i.text for i in a_token.children]
[]

# Find leftmost children
>>> [i.text for i in spacy_token.lefts]
[]
>>> [i.text for i in is_token.lefts]
['Spacy']
>>> [i.text for i in a_token.lefts]
[]

# Find rightmost children
>>> [i.text for i in spacy_token.rights]
[]
>>> [i.text for i in is_token.rights]
['tool', 'and', 'want']
>>> [i.text for i in a_token.rights]
[]

# Find all children
>>> [i.text for i in spacy_token.subtree]
['Spacy']
>>> [i.text for i in is_token.subtree]
['Spacy', 'is', 'a', 'great', 'tool', 'and',
'I', 'want', 'to', 'learn', 'more', '.']
>>> [i.text for i in a_token.subtree]
['a']

References

DependencyParser Class

Part-of-Speech Tagging

Named Entity Recognition