3. Part-Of-Speech Tagging¶
A Part-of-Speech (PoS) is a category of words, that have similar grammatical properties. A simplified set of Part-of-Speeches contains nouns, verbs, adjectives, adverbs, determiners, etc. For some tasks more detailed categories such as nouns-singular, nouns-plural, proper noun, etc. are applied.
In NLP the process of assigning PoS to the given words in a sentence is called PoS tagging. Depending of the word’s definition and it’s context (the surrounding words) a PoS tagger tries to assign the correct PoS to each word in the given text. PoS-tagging is challenging because many words can have different PoSs, e.g. run, love, …
Knowing the PoS of the words is of benefit for many applications, e.g.
once the PoSs of all words in a sentence are known, a syntax parser can calculate the syntactic tree of the sentence and it can determine if the syntactic structure is correct (syntax correction in editors)
from the PoS and the position within the syntax-tree the role/function of the word can be and derived, which is a key input form semantic analysis
in some NLP applications only words belonging to certain PoSs are required. E.g. for sentiment-analysis adjectives and adverbs are informative, but not determiners.
In this section common PoS and Pos-Tagsets are introduced as well as algorithms for PoS-Tagging. Moreover, it is demonstrated how packages like TextBlob or spaCy can be applied for PoS-Tagging.