Text Classification

Supervised Learning

Notes:

What?

Assign document to class(es)
Based on document features:
- Contents
- Age
- Popularity
- …

Notes:

What are use cases?

Why?

Standing queries (Google Alerts)
Spam filter
Webspam detection
Language detection
Sentiment detection

Notes:

How?

Make humans do it
- Needs domain experts
- Does not scale with amount of data
Make machines do it
- Needs lots of data

Notes:

How?
What's the issue with humans doing it?

Supervised learning

Notes:

Supervised Learning Example

Notes:

Nomenclature

Learning Method: Computes classifier
Classifier: Determines class of document

Class membership

Type	Example
Single / one-of	Spam
Multiple / any-of	Blog categories

Classifiers

Type	Example
One class	Spam
Two class	Sentiment
Multi class	Blog categories

Notes: Examples?