Scribd
Upload
5 views3 pages

POS Tagging Comparison

Uploaded by

bharathmad30
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
5 views3 pages

POS Tagging Comparison

Uploaded by

bharathmad30
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Part of Speech (POS) Tagging

POS tagging is the process of assigning a part of speech (e.g., noun, verb, adjective)
to each word in a sentence based on its definition and context.

1. Rule-Based POS Tagging


Definition:
Rule-Based POS tagging uses a set of hand-crafted linguistic rules to assign tags to
words.

Process:
Relies on lexical dictionaries (e.g., "run" can be a noun or verb).
Uses rules to resolve ambiguities.
Example rules:
If a word ends in “-ly”, tag as an adverb (RB).
If a word is preceded by a determiner (e.g., “the”) and not a verb, tag as a
noun.

Example:
Sentence: “He can fish.”
“can” → modal verb (MD)
“fish” → verb (VB) based on rule that a modal is followed by a verb.

Strengths:
Linguistically interpretable and explainable.
Effective when the language is well-understood.
No training data required.

Weaknesses:
Difficult and time-consuming to write exhaustive rules.
Not robust to exceptions and ambiguous contexts.
Poor performance in noisy or unseen data.

2. Stochastic (Statistical) POS Tagging


Definition:
Stochastic POS tagging uses probabilities and statistics from a large corpus to
determine the most likely tag for a word in context.

Types:
Unigram Tagger: Uses most frequent tag for each word.
Bigram/Trigram Tagger: Uses previous one or two tags to determine the current
tag.
HMM Tagger: Uses Hidden Markov Models for sequence-based tagging.
Machine Learning Models: Use Naïve Bayes, CRFs, or Neural Networks.

Example:
Sentence: “He can fish.”
The model calculates probabilities:
P(can|PRP) = modal verb (MD)
P(fish|MD) = verb (VB)

Strengths:
Learns from data; adaptable to new patterns.
High accuracy with large, annotated corpora.
Handles ambiguity statistically.

Weaknesses:
Requires large labeled datasets for training.
Less interpretable than rule-based methods.
Performance depends on training corpus quality.

Comparison Table:
Criteria Rule-Based Tagging Stochastic Tagging

Approach Manual rules Probabilistic models

Data Requirement No training data needed Requires annotated

Adaptability Low High

Accuracy (in practice) Moderate High (with sufficient

Interpretability High (transparent rules) Low (black-box for

Example of Ambiguity Hard to resolve all cases Uses probability to

Conclusion:
Rule-based tagging is best for small systems or where rules are well-defined.
Stochastic tagging is more scalable and accurate for large, real-world
applications.

You might also like