Text Classification and Sequence Labeling in NLP

April 10, 2021 2 minute read

In this blog I am describing a short note about what is text classificaiton and sequence labeling in NLP.

Text Classification

Text classification measns classify a sentence with it’s corresponding label.

In short:

Given a sentence X and predict an output Y

Topic Classification

Topic classifcation predict a topic(like food, sports, music) from an input sentence or text

For Example:
```
I like pizza -> food
I like Ariana Grande -> music
------------------------
```
Sentiment Analysis

Sentiment analysis predict a sentiment (like positive, negative or neutral) from an input sentence or text

For example:
```
I like this picture -> positive
I hate this picture -> negative
  
```
Language Identification

Language identification predict a language type(like English, Bangla…..) from an input sentence or text

For Example:
```
She live in Dhaka -> English
সে ঢাকায় বাস করে -> Bangla
```
Hate Speech Detection

Hate speech detection detect hateness (like political, religious, personal, race….) from an input sentence or text

For example:
```
He is ugly -> personal
```
and there are so many classification problem arround here

Sequnce labeling means generate a sequence of label from an input sentence

In short:

Given an input sentence X and generate sequence of label Y of equal length

Part of Speech Tagging

Part of speech tagging problem generate different part of speech tag from an input sentence

For example:
```
I eat rice -> PRON VERB NOUN 
```
Lemmatization

Lemmatization predict different lemma from an input sentence

For example:
```
He ate rice -> He eat rice
```
Language Identification

Language identification predict different language type from an input sentence

For example:
```
ami tomake love kori -> BN BN EN BN
```
and there are more sequence tagging problem around here

Span labeling kind of sequence labeling but instead of labeling sequence it’s label span by speical kind of sequence label

Name Entity Recognition

Name entity recogtion predict different name entity(like PERSON, LOCATION) from an input sentence

For Example:

Sagor Sarker love to visit Rangpur -> B-PER I-PER O O O S-LOC -> (Sagor Sarker)-> PERSON, Rangpur->LOCATION
  

you can find different type of name entity tagging format here

More example for other type sequence labeling:

semantic role labeling

Sagor Sarker is living in Dhaka, Bangladesh -> (Sagor Sarker)-> Actor, living->Predicate, (Dhaka, Bangladesh)->Location

syntactic chunking

Sagor Sarker is living in Dhaka -> (Sagor Sarker)->NP, (is living)-> VP, (in Dhaka)->NP