Analysis of Twitter Data

14
Analysis of Twitter Data NIKHIL PURANIK CMSC 601 – Research Skills 25 th April 2011 UNIVERSITY OF MARYLAND BALTIMORE COUNTY

description

UNIVERSITY OF MARYLAND BALTIMORE COUNTY. 25 th April 2011. CMSC 601 – Research Skills. Analysis of Twitter Data. NIKHIL PURANIK. Image adopted from slides by Akshay Java. A Bird’s Eye View: Twitter is a micro blogging website. - PowerPoint PPT Presentation

Transcript of Analysis of Twitter Data

Page 1: Analysis of Twitter Data

Analysis of Twitter Data

NIKHIL PURANIK

CMSC 601 – Research Skills

25th April 2011UNIVERSITY OF MARYLAND BALTIMORE COUNTY

Page 2: Analysis of Twitter Data

Image adopted from slides by Akshay Java

Page 3: Analysis of Twitter Data

A Bird’s Eye View:

Twitter is a micro blogging website.

Used by millions of people all over the world

A big repository for real time information

Twitter asks the question :“what is happening?” The user enters the information in 140 characters or less.

This information can be anything , what is on user’s mind, opinions about a topic, political views, reactions about events.

Page 4: Analysis of Twitter Data
Page 5: Analysis of Twitter Data

Tweets – Are they really useful??

Page 6: Analysis of Twitter Data
Page 7: Analysis of Twitter Data
Page 8: Analysis of Twitter Data

Twitter Facebook

No limits

D.O.B required

Easy for Ads.

No Profile info.

140 characters

Page 9: Analysis of Twitter Data
Page 10: Analysis of Twitter Data
Page 11: Analysis of Twitter Data

Research Problem Addressed/Topic of Interest

Twitter does not ask a user for any kind of personal information

Language to tweet in, time zone, tweet location, tweet media and Bio (160 char.)

Having information such as education, philosophy, sports, activities and interests, relationship status, etc. beneficial

Providing selective and focused advertisements

Page 12: Analysis of Twitter Data

Proposed Approach

Collection of Tweets

Extracting the features

Machine Learning for classification purpose

Controlled Experiments – Amazon’s Mechanical Turk

Testing on actual Tweets

Page 13: Analysis of Twitter Data

Christopher Horn-Analysis and Classification of Twitter

messagesGraz University of

Technology

• Classification of a tweet(and user) into news, user and company and fact/opinion

• 80% accuracy with SVMs

Mike Thelwall et al. -

Sentiment in Twitter Events

• The sentiment for each tweet was classified using an existing

• algorithm named SentiStrength which classifies a sentiment as positive or negative

Related work :

Page 14: Analysis of Twitter Data

Thank you!