social media, linked'data & the context question
DESCRIPTION
Presented by Professor Pete EdwardsTRANSCRIPT
Social Media, Linked Data & The Context Ques9on
Pete Edwards [email protected]
Compu'ng Science / dot.rural Digital
Economy Hub University of Aberdeen
dot.rural Research Hub • Exploring the contribu1on digital
technologies can make to enhancing key services ; genera1ng business opportuni1es ; boos1ng quality of life ; promo1ng the economic, social and environmental sustainability of rural areas across the UK and beyond.
• Researchers from a range of disciplines: • computer science,
communica'ons engineering, human geography, sociology, environmental science, business, medicine, transport studies
Accessibility & Mobilities
Enterprise & Culture
Natural Resource
Conservation
Healthcare CS / Eng
Agenda
• Social Media – PiFalls • The Context Challenge • Provenance • Digital Social Research Experience • Quality • Open Ques'ons
Social Media – PiFalls #1
• Integra'ng different informa'on sources to support public transport informa'on – Government open data (na'onal and local), operator data, vehicle data (when available), disrup'on reports …
• “Can we use TwiUer data?” – OXen used as a means to report disrup'on/service issues.
UK Snow – January 2013 “Snow causing #chaos in Cheltenham! Reports of up to 1.5 inches in worst affected areas... #uksnowdisaster!”
“No trains between Portsmouth Harbour & Southampton Central un9l further no9ce #uksnow #SnowSouthern” – SouthernRailUK
“Avoid the Penn road. It is awful!!! #uksnow #wolves #wolvessnow”
How to assess veracity of such reports?
Social Media – PiFalls #2 • Big Data: PiDalls, Methods and Concepts for an Emergent Field,
Zeynep Tufekci 2013 Ø Use of social media analysis by social scien'sts and policy makers is challenged by inadequate aUen'on to methodological and conceptual issues.
• Issues: 1. LiUle/no aUen'on to the implicit and explicit structural biases of the
plaForm(s) most frequently used to generate datasets. 2. Lack of clarity with regard to sampling, universe and
representa'veness (who are the ‘crowd’?). 3. Most analyses come from a single plaForm.
The Context Challenge • Integra1on of social media data with other datasets needs to address issues around data quality.
• Social media analyses need to be more explicit in discussing the biases, implicit assump1ons and steps in analy1cal methods.
• How do we understand the origins of data, to help assess veracity and u'lity?
• How do we document the processes involved in analysis?
• How do we integrate different sources together?
Provenance
• Lineage, history, audit trail… • Who, What, Where, Why, When, Which, & (W)How (Goble, 2002).
• W3C Provenance Working Group – “Informa'on about en''es, ac'vi'es, and agents involved in producing a piece of data or thing, which can be used form assessments about its quality, reliability, or trustworthiness”
W3C PROV Model
W3C PROV hUp://www.w3.org/TR/prov-‐overview/ hUp://www.w3.org/TR/2013/REC-‐prov-‐o-‐20130430/
W3C PROV Model
• En'ty – “a physical, digital, or other kind of thing with some fixed aspects”
• Ac'vity – “something that occurs over a period of 'me and acts upon or with en''es”
• Agent – “something that bears some form of responsibility for an ac'vity taking place, the existence of an en'ty, or another agent’s ac'vity”
Annota'on: “Angry”
Sen'ment classifica'on
Classifer Service
wasGeneratedBy
wasAssociatedWith
Linked Data
• Using the Web to connect related data that wasn’t previously linked
• Designed for humans and machines. • Links between a thing and its descrip'on
– RDF (Resource Descrip'on Framework) – Pete -‐> works For -‐> University of Aberdeen
• Encourages reuse, reduce redundancy.
Digital Social Research Experience
Quality
• Reasoning about quality seen as critical as more and more services/things publish data.
• Quality – a measures of ‘fitness for use’. • Quality metrics should examine the context
around data (including provenance). – Outputs are quality scores categorised into quality
dimensions (e.g. accuracy, relevance, …)
Qual-‐O
Assessment
Metric
Subject
ResultresultOf
guidedBy
assesses
AssessmentAssessment_1
MetricAccuracyMetric_1
SubjectObservation_2
ResultAccuracyResult_1resultOf
guidedBy
assesses
ObservationObservation_2
SensorOutputSensorOutput_2ssn:observationResult ssh:hasValue
"10"sensors:error
ObservationValue
OResult_2
spin:rule
Assessment
MetricAccuracyMetric_1
Subject
ResultresultOf
guidedBy
assesses
spin:rule
CONSTRUCT { […]} WHERE { ?pObs ssn:observationResult ?so . ?so ssn:hasValue ?ov . ?ov sensors:error ?error . BIND ((1 - (?error / 50)) AS ?qs) . […]}
Qual-‐O
Entity
AssessmentAssessment_2
MetricAccuracyMetric_2
SubjectObservation_2
ResultAccuracyResult_2resultOf
guidedBy
assesses
ObservationObservation_2
spin:rule
EntityObservationObservation_1
wasDerivedFrom
"50"sensors:error
ObservationValue
OResult_1
AssessmentAssessment_2
MetricAccuracyMetric_2
SubjectObservation_2
ResultAccuracyResult_2resultOf
guidedBy
assesses
spin:rule
AssessmentAssessment_2
MetricAccuracyMetric_2
SubjectObservation_2
ResultAccuracyResult_2resultOf
guidedBy
assesses
CONSTRUCT { […]} WHERE { ?obs (prov:wasDerivedFrom)+ ?pObs . ?pObs ssn:observationResult ?so . ?so ssn:hasValue ?ov . ?ov sensors:error ?error . BIND ((1 - (?error / 50)) AS ?qs) . […]}
Ques'ons
• How should provenance aUributes and characteris'cs be defined for social media?
• Can provenance be iden'fied and leveraged to help influence analysis?
• In addi'on to veracity, what other connec'ons can be made between provenance and elements of social media?
• What are the implica'ons for privacy?
TwiUer Ethnography 16-Jan Aberdeen Operator FirstAberdeen Morning Aberdeen, our control room team are
reporting full service out there today, with only minor delays at the moment
1.01
16-Jan Aberdeen Public mikewareham @FirstAberdeen just realised your driver gave me a single ticket this morning when I asked for, and paid for, a return! Ridiculous!
2.01
16-Jan Aberdeen Operator FirstAberdeen @mikewareham Hi Mike, sorry about that - can you email us the details to [email protected] and we'll investigate. Thanks
2.02
16-Jan Aberdeen Public dalgarnoamanda @FirstAberdeen thanks to the no3 bus driver that's waited for me three days in a row while I run for the bus. Much appreciated!
3.01
16-Jan Aberdeen Operator FirstAberdeen @dalgarnoamanda You're welcome - do you have a bus number so that we can pass on your thanks to the driver?
3.02
16-Jan Aberdeen Public dalgarnoamanda @FirstAberdeen its the no3 that is on rosemount at the moment going towards town
3.03
16-Jan Aberdeen Operator FirstAberdeen @dalgarnoamanda That bus isn't tracking at the moment, but the bus number's on your ticket - should be 5 digits starting with a 6
3.04
16-Jan Aberdeen Public dalgarnoamanda @FirstAberdeen hold on its 62191 3.05 16-Jan Aberdeen Operator FirstAberdeen @dalgarnoamanda That's the one - thanks for that,
we'll pass on your compliments 3.06
17-Jan Aberdeen Operator FirstAberdeen It's Friday Aberdeen! Yay! 2 minor RTC's at Nigg and Mounthooly are causing some delays to services 18 and 11,20 & 23 at the moment
4.01
17-Jan Aberdeen Operator FirstAberdeen But apart from that, it's a quietish day on Aberdeen's roads with full service and only minor delays - let's hope it stays that way!
5.01
16-Jan Aberdeen Public AndrewWatt7 @FirstAberdeen is there a bus that go's from union square to pittordrie football ground?
6.01
17-Jan Aberdeen Operator FirstAberdeen @AndrewWatt7 Hi Andrew, Ser 13 will take you to the back of the ground, or a 1, 2 or X40 will drop you on King St and you can walk through
6.02
17-Jan Aberdeen Public AndrewWatt7 @FirstAberdeen okay thanks 6.03
16-Jan Aberdeen Public FraserMacaulay @FirstAberdeen where did you get your no.17 driver? The moon? Absolute ****
7.01
17-Jan Aberdeen Operator FirstAberdeen @FraserMacaulay Hi Fraser, can you email us the details to [email protected] and we'll have a word with the driver. Thanks
7.02
• Contextual inquiry • Digital diary study • Content analysis
Thanks …
• Team members: – Chris Baillie, David Corsar, Milan Markovic, Edoardo Pignon, Paul Gault
• Collaborators: – Caitlin CoUrill, John Nelson, Jillian Anable
• Partners:
• Funders: