# Sentiment analysis

![](Files/sent1.png)

Sentiment is multifaceted.
There are multiple levels of affective states

![](Files/sent2.png)

## Basic sentiment classification
![](Files/sent3.png)

Worked example
![](Files/sent4.png)

Two optimizations can be made
![](Files/sent5.png)

Occurrence is more important than frequency

![](Files/sent6.png)
We need to work out the negations

## Sentiment lexicons
When training data is not available or is limited for the goal of building models, we can use sentiment lexicons, which are manually edited word lists.

![](Files/sent7.png)

## Summary

![](Files/sent8.png)
![](Files/sent9.png)
![](Files/sent10.png)
![](Files/sent11.png)
![](Files/sent12.png)