Back to Module 8

Dependency parsing

Dependency parsing identifies grammatical relationships between words in a sentence by representing how individual words are connected to one another. (Note that I use dependency parsing as a general term for both the automatic assignment of dependency relations to words (dependency tagging) and the analysis of the resulting dependency structure. This is because, in practice, dependency analysis involves both identifying dependency labels and examining the head–dependent relationships they represent.)

Unlike POS tagging, which assigns a grammatical category to each individual token, dependency parsing describes the grammatical relationships between tokens.

Consider the sentence:

Students analyze data.

A dependency parser may represent the sentence as:

Token UPOS Dependency relation Head
Students NOUN nsubj analyze
analyze VERB ROOT analyze
data NOUN obj analyze

Here:

  • Students is the subject (nsubj) of analyze;
  • data is the object (obj) of analyze;
  • analyze is the root of the sentence.

Consider another example:

Researchers carefully analyze complex data.

A dependency parser may identify relationships such as:

Token UPOS Dependency relation Head
Researchers NOUN nsubj analyze
carefully ADV advmod analyze
analyze VERB ROOT analyze
complex ADJ amod data
data NOUN obj analyze

This representation tells us not only what grammatical category each word belongs to, but also how the words are related. For example:

  • Researchers functions as the subject of analyze;
  • carefully modifies analyze;
  • data functions as the object of analyze;
  • complex modifies data.

Heads and dependents

A dependency structure consists of relationships between heads and dependents. Each dependent is connected to a head through a particular dependency relation.

For example:

complex sentence

The adjective complex modifies the noun sentence.

complex --amod--> sentence

Further reading

Next: English dependency parsing