mining climate change awareness on twitter: a pagerank network analysis method: iccsa'15
TRANSCRIPT
Mining Climate Change Awareness on Twitter
University of Vermont EPSCOR. Burlington VT, USA
Ahmed Abdeen Hamed, Ph.D.
Asim Zia, Ph.D.
Motivation
Public opinions are powerful tools
Identifications of weaknesses and strengths
Better educating for the general public
Understanding response to decisions
Making new policies and decisions
http://thebridge.agu.org/thebridge/files/2013/10/pollinggraph.gif
In the UK, Leorenzoni et al. studied how public perceives Climate ChangeIn Japan, Sampie et al. analyzed newspapers for public opinion on Global WarmingIn the US, Semenza et al., designed surveys to compare individuals live in Portland OR vs Houston, TX
Background Research
Why Twitter?
The place for honest, short, to the point opinions, scary
Very large content connected to seasons, events
Easy APIs which makes it attractive to mine
Various content devices to measure Words Hashtags Links Impressions
Mining For
Climate Change
We need a set of terms that highly represent
Climate to search Twitter for
We need to rank the terms according to their
importance
We need an objective methods of ranking
We need a notion of ground-truth to compare
against
Ground-Truth
A set of terms :
Carefully selected (TF-IDF)
Closely related (Bigrams)
From an authentic source (Nature Climate Change)
Twitter Bigram Network
Using the TF-IDF Features from Nature Climate Change
Searched 72 million tweets dataset
Constructed Twitter Bigram Network (TBN)
Weights assigned based on the frequencies of bigram
Measured PageRank for each term
PageRank Network Analysis
For each term in the GTN we measured its PageRank
For each term in the TBN we measured its PageRank
Compared Ranks of the same term in the two networks
Comparison Results
TBN
• Generic terms (bigger size)
• Not directly related
• Some related terms (smaller size)
Insights
Words are not the only textual features
The social language differs from the academic
Hashtags are very common on Twitter/Facebook
Social impressions (Likes, RT, Shares, Replies, Mentions were not considered)
Sentiments and emoticons were not considered
Investigating Hashtags
Searching tweets for hashtags composed of bigrams
Performing Association Analysis experiments
Using Confidence as a measure of awareness
Conclusions
There is indeed climate change awareness on Twitter
#ClimateChange hashtag #1
Compound hashtags are highly ranked thank single
Still there are missing terms from the big picture #Carbon #CarbonOffsets
Conclusion and Future
There is much awareness on Twitter
Awareness is general but not technical
Hashtags represent a very important element in measuring public awareness
Future: Studying events Combining Sentiments Time series analysis
Acknowledgement
This research is funded by Vermont EPSCoR Award EPS-1101713
Authors thanks the following for their efforts: Alexa Ayer of UVM Rubenstein collecting the
data Dr. Ilan Kelman for the valuable discussions Dr. Ibrahim Mohamed for the visualization and
analysis insights