using social media for ecosystem observations: practical approaches, analyses and how-to tips for...
TRANSCRIPT
Jeremiah Osborne-‐Gowey, Feather River Consul:ng, @JeremiahOsGo Mary Fuka, EnPhysica LLC, @MzPhyz Daniel Fuka, Dept. of Biological Systems Engineering, Virginia Tech, M. Todd Walter, Biological and Environmental Engineering, Cornell University Zachary Easton, Dept. of Biological Systems Engineering, Virginia Tech
Using Social Media for Ecosystem ObservaJons
PracJcal Approaches, Resources And Tips for ScienJsts and Resource Managers
Big data, small tools, huge insights
• Lots of free data out there, waiJng
• Free tools to help
Digital Age
InformaJon processing
Digital Age
Interconnected networks
DisJll to relevant data. Find paOerns.
InformaJon Overload?
‘Big data’ black hole
‘Big data’ black hole
InformaJon fire hose
‘Big data’ black hole
InformaJon fire hose
‘Big data’ black hole
InformaJon fire hose
It’s not informaJon overload. It’s filter failure. ~Clay Shirky @cshirky
Twi>er ❧ What are you
doing? ~140chars
Twi>er ❧ What are you
doing? ~140chars ❧ 1 Billion+ total
users, 255 million monthly active users
❧ 50 million tweets sent every day
❧ 46% of users tweet 1+ times a day
❧ 19% of all adults online are on Twitter
Twi>er ❧ What are you
doing? ~140chars ❧ 1 Billion+ total
users, 255 million monthly active users
❧ 50 million tweets sent every day
❧ 46% of users tweet 1+ times a day
❧ 19% of all adults online are on Twitter
Twi>er ❧ What are you
doing? ~140chars ❧ 1 Billion+ total
users, 255 million monthly active users
❧ 50 million tweets sent every day
❧ 46% of users tweet 1+ times a day
❧ 19% of all adults online are on Twitter
Twi>er ❧ What are you
doing? ~140chars ❧ 1 Billion+ total
users, 255 million monthly active users
❧ 50 million tweets sent every day
❧ 46% of users tweet 1+ times a day
❧ 19% of all adults online are on Twitter
TIP: Use TwiOer’s naJve search to find data of interest. Be creaJve. But be sure to check the “All” tweets link, not just “Top Posts”
HOT TIP: Develop species/topical keyword list, get this book/codes
TwiOer: @SocialWebMining @ptwobrussel
HOT TIP: Examine raw (TwiOer) data for specific ‘sighJng’ language (e.g., caught, hooked, landed, saw, hit, there is/are, OMG, etc.), then filter on those word associaJons
0
100
200
300
400
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
cherry_reports
0.0
2.5
5.0
7.5
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
chokecherry_reports
0
100
200
300
400
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
dogwood_reports
0
50
100
150
200
250
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
armadillo_reports
Examine filtered data
What info can be extracted? Location Date/Time Environment Reaction/Sentiment Behavior Condition
What info can be extracted? Date/Time
LocationEnvironment
Reaction/Sentiment Behavior Condition
Profile informationConnections…
SEASONALITY
0
10
20
30
40
50
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
hummingbird_reportsHUMMINGBIRD -‐ DAILY
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●●
● ●
●●
●
●
●
●
● ●
●
●● ●
●
●
●
●
●
●● ●
●●
●●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●●
●●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
● ●
●●
●
●
●●●
● ●
●●●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●●
●
●● ●
●
●
●
●●
●
●
●
●●
●
●
●
●●
●●●
●
●
●●●
●0
20
40
60
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
dogwood_reportsDOGWOOD -‐ DAILY
0
25
50
75
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
cherry_reportsCHERRY -‐ DAILY
ANGLER EFFORT/ RECRUITMENT
IDENTIFY/EARLY TRACKING OF INVASIONS, RANGE SHIFTS
0
10
20
30
40
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
armadillo_reportsARMADILLO -‐ DAILY
HOT TIP: Expect 2-‐9% of tweets to come with geolocaJons. A few other tweaks can drive that number closer to 40%.
TRACK DISEASE VECTORS, OUTBREAKS
SUPPLEMENT EXISTING ECOSYSTEMS INFO & DATABASES
STEELHEAD 2/7-‐3/31
OPINION MINING / SENTIMENT
SENTIMENTVIZ (CHRIS HEALY)
HOT TIP: Check out Chris Healey’s (NCS) SenJmentViz, SenJWordNet for quick, useful opinion mining of TwiOer.
2014 urban wildlife hotspots FACTOID: 85% of US lives in urban areas. 16% use Twitter. It’s wild what they tweet.
Take Home Messages
• Networking plakorms freely provide access to lots of data
• Useful ecosystem observaJons can be mined from social datastreams
• There are a number of free tools to help find, store and filter social-‐derived data
Jeremiah Osborne-‐Gowey Feather River ConsulJng Contact: @JeremiahOsGo
Zachary Easton Dept. of Biological Systems Engineering
Virginia Tech
Mary Fuka EnPhysica, LLC @MzPhyz
Daniel Fuka, Dept. of Biological Systems Engineering
Virginia Tech
M. Todd Walter Biological and Environmental Engineering Cornell University
πάντα χωρεῖ καὶ οὐδὲν μένει • Heraclitus of Ephesus (535-‐475BC)
Everything changes and nothing stands sJll. • Heraclitus of Ephesus (535-‐475BC)
InformaJon bombardment.
0
5000
10000
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
all_cherry
0
25
50
75
100
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
all_chokecherry
0
2000
4000
6000
2011−Nov 2012−Feb 2012−May 2012−Aug 2012−Nov 2013−Feb 2013−May
Tweets
all_dogwood