Constructing Social Media Knowledge Graphs with Social Scientists
1 like895 views
This document discusses constructing knowledge graphs to help social scientists analyze social media data. It summarizes interviews with 12 social scientists who face challenges using existing tools for social media capture and analysis. The tools lack support for qualitative analysis and user-driven thematic coding. The document proposes using a knowledge graph approach to establish common ground between social scientists and tools by representing social media data as interconnected entities and attributes. Two prototype tools are described that allow thematic coding of data and visualization of analyzed data as knowledge graphs.
Constructing Social Media Knowledge Graphs with Social Scientists
- 1. Construc)ng Social Media Knowledge Graphs with Social Scien)sts John Paul Vargheese, Peter Travers, Je鍖 Z. Pan, Kathryn Vincent, Claire Wallace, Anna Kabedeva
- 3. Mo)va)on ≒ Signi鍖cant opportuni?es for social scien?sts to discover novel aspects of human behaviour ≒ Social interac?ons ≒ Social and personal expressions ≒ Poli?cal expressions, commentaries and alignments
- 4. Key ques)on 1: Are exis)ng tools ready for interdisciplinary research on social media? ≒ Increasing number of tools to support so media research have emerged ≒ 230 tools listed on the Social media monitoring wiki (hUp://wiki.kenburbary.com/social-meda-monitoring-w
- 5. Common Features among exis)ng tools ≒ Capture / Data collec?on ≒ Sta?s?cal analysis ≒ Natural language processing ≒ Sen?ment analysis ≒ Argument mining ≒ User de鍖ned thema?c coding and annota?on of social media data ≒ Varying forms of visualisa?on op?ons to present analysis
- 6. Semi-structured interviews ≒12 par?cipants are social scien?sts currently working with social media data ≒Background of the par?cipants ≒ 4 sociologists ≒ 3 Human geographers ≒ 1 design ethnographer ≒ 1 Poli?cal scien?st ≒ 1 Educa?on ≒ 1 Informa?on science ≒ 1 Humani?es (media studies Womens engagement in the media) Credit: hUp://www.f
- 8. Challenges for social scien)sts Exis)ng tools ≒ Not always accessible for social scien?sts ≒ Increasing mone?sa?on of social media ≒ Concern over storage, ethical and legisla?ve implica?ons ≒ Increasing costs of the tools available for capturing and analysis social media ≒ Lack of awareness for exis?ng tools and analy?cal frameworks concerning social media ≒ Nega?ve percep?on of such tools amongst social scien?sts ≒ Lacking re鍖ec?vity required for interpreta?on ≒ Insu鍖cient support for a more qualita?ve analysis ≒ Means of capture con鍖ic?ng with social science theore?cal and methodological approach
- 9. Nega)ve percep)on of automated analysis ≒Yes, the machine will do a wonderful job of coun7ng ... So we know how many followers you have, we know how many people had this hashtag. Did we recognise the sarcasm? No, we didnt. Par?cipant 11 ≒ I think just because I feel it doesn't pick up on sarcasm and irony and so on. And I think also as well with the more recent study I did, I went away from the elec7on campaign and looked at how siGng MSPs were using TwiKer, par7cularly for providing informa7on to their local cons7tuents and I think you really needed that manual, the human knowledge. Par?cipant 7
- 10. Observa)ons ≒Capture is the main priority ≒ Exis?ng means of capture are ohen unsuitable e.g. search, copy and paste ≒ Time consuming ≒ Unreliable ≒More support for qualita?ve means of analysis ≒ User driven thema?c analysis ≒Lack of mutual understanding between machine and users
- 11. Key ques)on 2: how to establish some common ground between social scien)sts and tools? ≒ Increasing number of tools to support so media research have emerged ≒ 230 tools listed on the Social media monitoring wiki (hUp://wiki.kenburbary.com/social-meda-monitoring-w
- 12. Knowledge Graph: What and Why? ≒A knowledge graph is a set of interconnected typed en??es and their aUributes ≒ based on a knowledge representa?on approach called seman?c network ≒ Standards (RDF, OWL, SPARQL etc.) and tools established in the Seman?c Web community ≒Used by Google from 2012 ≒ allowing users to search for things, people or places ≒ rather than just matching strings in the search queries with those in web documents 12
- 15. Knowledge Graph - Example
- 16. Knowledge Graph: What and Why? ≒Deriving meaningful facts by iden?fying rela?ons amongst rich data aUributes such as followers, friends, likes ≒How can we use this approach to support social scien?sts analysis of social media?
- 18. ≒A social scien?st is inves?ga?ng how narra?ve are formed by YES campaigners ≒The social scien?sts uses our tool to collect a data set of all Tweets posted by the campaigners ≒Example en??es include: users, tweets, loca?on, groups, topics, events Example: Understanding TwiNer data on SI-Ref Overview of the process
- 19. Discussions and Outlook ≒ Our interviews suggest there is a gap between social scien?sts and compu?ng science tools ≒ Idea: use Knowledge Graph as a common ground ≒ On-going and future work ≒ Evalua?on studies with social scien?sts ≒ Observa?onal studies involving using our tool ≒ Follow up interviews to re鍖ne preliminary requirements ≒ Incorpora?ng more sophis?cated means of analysis to further extend the reasoning driving the knowledge graph ≒ Structuring the knowledge graph using theore?cal approaches towards analysing social media such as the Honeycomb framework (Kietzmann et al. 2011)
- 21. Further Reading on Knowledge Graphs ... Je鍖 Z. Pan, Guido Vetere, Jose Manuel Gomez Perez and Honghan Wu (Eds.). Exploi?ng Linked Data and Knowledge Graphs for Large Organisa?ons. Springer. 2016.
- 22. More on Knowledge Graphs ... The 12th Interna2onal Reasoning Web Summer School, 5-9 Sept, 2016. Aberdeen, with limited posi2ons available. The 10th Interna2onal Conference on Web Reasoning and Rule System, 9-11 Sept, 2016. Aberdeen, UK.
- 23. Construc)ng Social Media Knowledge Graphs with Social Scien)sts Thank you! jeff.z.pan@abdn.ac.uk homepages.abdn.ac.uk/jeff.z.pan/pages/ @jpansw