dependency_parse

Analyze the grammatical structure and dependencies between words.

!!! note "Requires Deep Learning" This function requires the deep learning dependencies:

pip install "underthesea[deep]"

Usage

from underthesea import dependency_parse

text = "Tối 29/11, Việt Nam thêm 2 ca mắc Covid-19"
result = dependency_parse(text)
print(result)
# [('Tối', 5, 'obl:tmod'),
#  ('29/11', 1, 'flat:date'),
#  (',', 1, 'punct'),
#  ('Việt Nam', 5, 'nsubj'),
#  ('thêm', 0, 'root'),
#  ('2', 7, 'nummod'),
#  ('ca', 5, 'obj'),
#  ('mắc', 7, 'nmod'),
#  ('Covid-19', 8, 'nummod')]

Function Signature

def dependency_parse(sentence: str) -> list[tuple[str, int, str]]

Parameters

Parameter	Type	Description
sentence	str	The input text to parse

Returns

Type	Description
list[tuple[str, int, str]]	List of (word, head_index, relation) tuples

Each tuple contains:

• word: The word token
• head_index: Index of the head word (0 = root)
• relation: The dependency relation type

Dependency Relations

Relation	Description
root	Root of the sentence
nsubj	Nominal subject
obj	Object
obl	Oblique nominal
obl:tmod	Temporal modifier
amod	Adjectival modifier
nmod	Nominal modifier
nummod	Numeric modifier
punct	Punctuation
flat:date	Flat date expression
compound	Compound word

Examples

Basic Usage

from underthesea import dependency_parse

text = "Tối 29/11, Việt Nam thêm 2 ca mắc Covid-19"
result = dependency_parse(text)

for i, (word, head, rel) in enumerate(result, 1):
    print(f"{i}\t{word}\t{head}\t{rel}")
# 1    Tối         5    obl:tmod
# 2    29/11       1    flat:date
# 3    ,           1    punct
# 4    Việt Nam    5    nsubj
# 5    thêm        0    root
# 6    2           7    nummod
# 7    ca          5    obj
# 8    mắc         7    nmod
# 9    Covid-19    8    nummod

Finding the Root

text = "Tôi yêu Việt Nam"
result = dependency_parse(text)

root = [(i, word) for i, (word, head, rel) in enumerate(result, 1) if rel == 'root']
print(f"Root: {root}")
# Root: [(2, 'yêu')]

Finding Subjects and Objects

text = "Sinh viên đọc sách ở thư viện"
result = dependency_parse(text)

subjects = [word for word, head, rel in result if rel == 'nsubj']
objects = [word for word, head, rel in result if rel == 'obj']

print(f"Subjects: {subjects}")
print(f"Objects: {objects}")

Notes

• This function uses a transformer-based model
• First call may take longer due to model loading
• Requires significant memory for the deep learning model