Week 13

28 April :  Sentiment Analysis

Subject:   Sentiment analysis, sometimes called opinion mining, attempts to extract from texts affective or subjective information from data. The kind we will look at here is a somewhat simple one: the automated extraction, classification and interpretation of sentiment from texts using some natural language processing techniques. It is one of the ways we might say that we can “read like a computer.” Sentiment can also be derived from image or even biometric data.  We will look at two examples.  The first is sentiment analysis of books from Project Gutenberg and then second from Twitter.

Format:  Class this week will have one meeting on Thursday and will be otherwise asynchronous.  Attendance will be counted via your participation in the forum in the form of “quick writing” (a few sentences in response to a prompt).

To read before class:

On Hashtag Tracking Tools
Dobson, Mining Sentiment (in Drive)
Gendered Language in Teaching Reviews (Schmidt)
Twitter Hashtags: A Guide to Finding and Using the Right Ones
Problems with the Syuzhet package (Swafford)

To watch before class: 

Data Lit (sentiment analysis) (starting 1:00) (4 mins)
Setting up a Twitter API developer account (9 mins)

Outline of this week’s synchronous session:

(0) housekeeping (extensions, next week’s plan, final assignment)
(1) notebook 1: Detecting Sentiment in texts, and in Project Gutenberg Texts
(2) notebook 2: Syuzhet package
(3) notebook 3: Twitter sentiment
(4) Gendered Language in Teaching Reviews dashboard

Notebook:

There are two notebooks in the class drive folder: Detecting Sentiment in Texts and TwitterSentiment.
There are desktop sentiment analysis programs such as LingMotif as well (I have never tried them).

Quick writing:

In this week’s Forum add a few lines about what kinds of materials you are persistently reading. By reading, it can be, but does not have to be, books or fiction, since much of what we read comes as a “feed” or a “stream” (and sometimes a live one). Pick one of those persistent forms of reading that you would like to analyze for sentiment. What do you think you would find? Who might already be reading it for sentiment? Why?  Is there something like RateMyProfessor.com shown in the above mentioned interactive visualization “Gendered Language” that you think would provide a great test case for sentiment analysis?