Time seriesTime seriesWord frequenciesWord frequenciesSentiment analysis Sentiment analysis

Twitter Analysis for the Social Sciences and Humanities

Twitter Data: Quick and Slow1. Quick: http://topsy.com search, then click

2. Slow: Download Mozdehhttp://mozdeh.wlv.ac.uk,collect data in advance

then analyse

Tweets per hourincluding ‘Nobel’

Tweets per dayincluding ‘Nobel’

Gathering Twitter data

Tweets can be obtained via Mozdeh Also: Chorus http://www.chorusanalytics.co.uk/

NodeXL and other software

Must be gathered in near to real time Not available after about 2 weeks

Gather using keyword or phrase searchesIf historical tweets needed, buy from a data provider

Twitter Time Series Analysis with Mozdeh: Step by Step

Download free from http://mozdeh.wlv.ac.uk for Windows

Search engine part of Mozdeh – for the saved tweets

Sentiment-based searches

Gender-based searches

Identifies words that are Relatively frequent for the query

Co-word analysis

Am automatic method to compare different subsets of tweets for key differencesIdentifies words that are relatively frequent compared to one keyword (or gender) compared to another

The above example will compare the words in tweets containing book and read by gender

Words more common in male tweets containing “book”than in female tweets containing “book”

Words more common in ? tweets than in ? Tweets(male or female)

Words more common in ? tweets than in ? Tweets(male or female)

A statistical measure of the significanceof the difference between genders

Twitter research ideas

Monitor keyword searches for a topicAnalysis ideas

Time series – identify trends & causes of any peaks

Content analysis of random sample tweets – why is the topic tweeted?

Gender/sentiment/high frequency keywords Network analysis of tweeters/tweeting &

qualitative analysis of key tweeters – who is tweeting and who is successful?

How and why scholars cite on Twitter – Priem & Costello

Example of a Twitter study using interviews and content analysis of tweetsGathered and analysed a sample of tweets sent by scholarsConcluded that: “Twitter citations are also uniquely conversational, reflecting a broader discussion crossing traditional disciplinary boundaries.”

High frequency word analysis

An analysis of high frequency words for Nobel prizes found tweets about: Alternative winners’ names non-

academic prizes only

Gender references Female winner only (9% mentioned her gender).

Expressions of Sentiment literature prize only (42% positive)

A quick analysis can give new insights.

Sentiment Strength Detection in the Social Web -SentiStrength

• Detect positive and negative sentiment strength in short informal text Develop workarounds for lack of standard

grammar and spelling Harness emotion expression forms unique

to MySpace or CMC (e.g., :-) or haaappppyyy!!!)

Classify simultaneously as positive 1-5 AND negative 1-5 sentiment

Thelwall, M., Buckley, K., & Paltoglou, G. (2012). Sentiment strength detection for the social Web.Journal of the American Society for Information Science and Technology, 63(1), 163-173.

Thelwall, M., Buckley, K., Paltoglou, G., Cai, D., & Kappas, A. (2010). Sentiment strength detection in short informal text.Journal of the American Society for Information Science and Technology, 61(12), 2544-2558.

SentiStrength Algorithm - Core

List of 2,489 positive and negative sentiment term stems and strengths (1 to 5), e.g. ache = -2, dislike = -3, hate=-4,

excruciating -5 encourage = 2, coolest = 3, lover = 4

Sentiment strength is highest in sentence; or highest sentence if multiple sentences

My legs ache.

You are the coolest.

I hate Paul but encourage him.

-2

3

-4 2

1, -2

positive, negative

3, -1

2, -4

Extra sentiment methodsspelling correction nicce -> nicebooster words alter strength very happynegating words flip emotions not nicerepeated letters boost sentiment/+ve niiiice emoticon list :) =+2exclamation marks count as +2 unless –ve hi!repeated punctuation boosts sentiment good!!!negative emotion ignored in questions u h8 me?Sentiment idiom list shock horror = -2

Online as http://sentistrength.wlv.ac.uk/

Tests against human coders

Data set

Positivescores -correlation with humans

Negativescores -correlation with humans

YouTube 0.589 0.521

MySpace 0.647 0.599

Twitter 0.541 0.499

Sports forum 0.567 0.541

Digg.com news 0.352 0.552

BBC forums 0.296 0.591

All 6 data sets 0.556 0.565

SentiStrengthagrees with

humansas much as theyagree with each

other 1 is perfect agreement, 0 is random agreement

Why the bad results for BBC? (and Digg)

Irony, sarcasm and expressive language e.g., David Cameron must be very happy

that I have lost my job. It is really interesting that David

Cameron and most of his ministers are millionaires.

Your argument is a joke.

$

Example – sentiment in major media events

Analysis of a corpus of 1 month of English Twitter posts (35 Million, from 2.7M accounts)Automatic detection of spikes (events)Assessment of whether sentiment changes during major media events

Automatically-identified Twitter spikes

9 Mar 2010

9 Feb 2010

Proportion of tweetsmentioning keyword

Thelwall, M., Buckley, K., & Paltoglou, G. (2011). Sentiment in Twitter events. Journal of the American Society for Information Science and Technology, 62(2), 406-418.

Chile

matc

hin

g p

ost

sSenti

ment

stre

ngth

Sub

j.

Increase in –ve sentiment strength

9 Feb 2010

9 Feb 2010

Date and time

Date and time

9 Mar 2010

9 Mar 2010

Av. +ve sentimentJust subj.

Av. -ve sentimentJust subj.

Proportion of tweetsmentioning Chile

#oscars%

matc

hin

g p

ost

sSenti

ment

stre

ngth

Sub

j.

Increase in –ve sentiment strength

Date and time

Date and time

9 Feb 2010

9 Feb 2010

9 Mar 2010

9 Mar 2010

Av. +ve sentimentJust subj.

Av. -ve sentimentJust subj.

Proportion of tweetsmentioning the Oscars

Sentiment and spikes

Statistical analysis of top 30 events: Strong evidence that higher volume

hours have stronger negative sentiment than lower volume hours

No evidence that higher volume hours have different positive sentiment strength than lower volume hours

=> Spikes are typified by small increases in negativity

SummaryTweets gathered free with Mozdeh Search tweets and conduct content analysis Time series analysis graphs – trends over time Identify topics causing high sentiment Identify differences by gender Identify important topics by high freq. keywords Identify important topic differences by co-words Pilot test your ideas first

Gather tweets in advance for a period of timeCan give quick insights into your research goals – or can be a primary research method