What is dependency parsing?
Share
Introduction
Natural Language Processing (NLP) is an interdisciplinary concept that uses synthetic intelligence and computational linguistics fundamentals to see how human languages interact with technology. There are many techniques used with NLP in which syntax and semantics are used for parsing. Parsing is the grammatical analysis of a sentence. For example, if a sentence is fed to NLP, parsing involves breaking the sentence into parts of speech, for example, noun, verb, etc.
Dependency parsing
Dependency parsing refers to examining the dependencies between the phrases of the sentence to determine the grammatical structure of a sentence. Dependency grammar provides a representation of a language as graphs. Nodes are words, and edges are dependencies. In this process, it is assumed that there is a direct relationship between each linguistic unit in a sentence. The relationships between each linguistic unit, or phrase, in the sentence, are expressed by directed arcs called dependency structures.
Dependency structure
Dependency structures are directed graphs that satisfy the following constraints:
- They have a single designated root node that has no incoming arcs.
- Each node has one incoming edge except the root node.
- There is a unique path to each node from the root node.
Example
In this example, we see that we can traverse from the root word score to any other word in the structure. If we need to move from "score" to "good", we will move from "score" to "marks" and then from "marks" to "good".
Dependency tags
An essential part of dependency parsing is the dependency tag. A dependency tag indicates the relationship between two phrases. It is the word that modifies the meaning of the other word. If we look at the above example, the word "Intelligent" is an adjective for the word "students". The dependency tag used here is amod, which stands for the adjectival modifier. The start of the arrow represents the parent, and the end represents the child or dependent.
Some of the common tags are used for the syntactical relationships between the words. They are described below:
Commonly Used Dependency Tags
Dependency Tag | Description | Dependency Tag | Description |
acl | clausal modifier of a noun (adnominal clause) | expl:pass | reflexive pronoun used in reflexive passive |
acl:relcl | relative clause modifier | expl:pv | reflexive clitic with an inherently reflexive verb |
advcl | adverbial clause modifier | fixed | fixed multiword expression |
amod | adjectival modifier | goeswith | goes with |
appos | appositional modifier | iobj | indirect object |
aux | auxiliary | list | list |
aux:pass | passive auxiliary | mark | marker |
case | case-marking | nmod | nominal modifier |
cc | coordinating conjunction | nmod:poss | possessive nominal modifier |
cc:preconj | preconjunct | nmod:tmod | temporal modifier |
ccomp | clausal complement | nsubj | nominal subject |
clf | classifier | nsubj:pass | passive nominal subject |
compound | compound | nummod | numeric modifier |
compound:prt | phrasal verb particle | obj | object |
conj | conjunct | obl:arg | oblique argument |
cop | copula | obl:lmod | locative modifier |
csubj | clausal subject | obl:tmod | temporal modifier |
dep | unspecified dependency | parataxis | parataxis |
det | determiner | punct | punctuation |
det:nummod | pronominal quantifier agreeing in the case with the noun | root | root |
det:poss | possessive determiner | vocative | vocative |
discourse | discourse | xcomp | open clausal complement |
expl | expletive |
Implementation
Now, let's create a dependency parser in Python. Dependency parsing in Python is very easy and straightforward. We need to install some libraries. The implementation of the code for the above example is given below:
import spacynlp=spacy.load('en_core_web_sm')text='Intelligent students score good marks easily.'for token in nlp(text):print(token.text,'->',token.dep_,'->',token.head.text)
Code explanation
- Lines 1–2: We import the Python library
spacyand then load thespacy pipelinefor supporting the English language. - Line 4: A sample text in the English language.
- Line 6: We have a Loop for each token in the text.
- Line 7: The
token.textreturns the token of a sentence,token.dep_returns the dependency tag for a word, and thetoken.head.textreturns the respective head word (to which the arrow is pointing).
