1. Michael Shuffett Virginia Tech Blacksburg, VA shuffett@cs.vt.edu
Twitter Metadata
Michael Shuffett Virginia Tech Blacksburg, VA
Primary Client: Mohamed Magdy,

2. Background + = Large number of tweet collections
CTRNet
IDEAL
QCRI
No collection level metadata +
No easy merging solution =
Poor collaboration support
Collections about same thing
These three collect information about events (among other things)

3. Project Goals Develop metadata standards for tweet collections
start, end timestamps
geographic coverage
details of how collection was prepared
Filtering
Cleaning
Enriching
Create software package that merges and describes multiple collections
Geographic coverage includes tweets, users tweeting
Details such as what keywords or hashtags were used, how data was filtered. Was

4. Open Provenance Model A P P1 P2 A1 A2 Ag used(R) wasTriggeredBy
wasDerivedFrom A P
used(R)
wasGeneratedBy(R) Ag
wasControlledBy(R)
Adapted from
Provenance is the process that lead to data

5. Challenges
Complete provenance is not typically provided in standard collection data
How to merge collections that come from any number of formats