Accountability Hack Redux
0 likes535 views
This document describes Parli-N-Grams, a tool for analyzing language used in political speeches from Hansard. It extracts N-grams (sequences of N words) from texts and allows users to search for related ideas. The tool can be used to analyze empathy, emotional reactions, and characteristics like hubris in speeches. Future work includes improving the tool, expanding N-gram searches, building profiles of MPs, and creating interfaces for both public and researchers.
Accountability Hack Redux
- 1. Parli-N-Grams Giuseppe Sollazzo @puntofisso Accountability Hack 2014 Redux Portcullis House, 15 January 2015
- 2. The language used in politics points to the political, cultural, social background of MPs, parties, and their time
- 3. A great search lets you discover knowledge by just searching
- 4. Parli-N-Grams A search and analysis tool for Hansard
- 5. An N-Gram is a sequence of N words extracted from a longer sentence N-Grams?
- 6. You can split any word sequence in n-grams lists e.g. “The quick brown fox” N-Grams?
- 7. “The quick brown fox” | 1-grams ● the ● quick ● brown ● fox N-Grams?
- 8. “The quick brown fox” | 2-grams ● the quick ● quick brown ● brown fox N-Grams?
- 9. “The quick brown fox” | 3-grams ● the quick brown ● quick brown fox N-Grams?
- 10. “The quick brown fox” | 4-grams ● the quick brown fox N-Grams?
- 11. 30” demo
- 15. What can we find out?
- 22. Related Ideas
- 23. 1. Create a Markov Chain by analysing text 2. Generate text by using the Markov Chain Markov Text Generation
- 24. photo CC BY-NC by UK Parliament: https://www.flickr.com/photos/uk_parliament/8719647747
- 25. What happens if we apply Markov analysis and generation to the Queen’s Speeches?
- 26. My Government will be devolved.
- 27. My Government will be devolved. FALSE
- 28. My Government will work with the Afghan government.
- 29. My Government will work with the Afghan government. TRUE
- 30. My Government looks forward to an enhanced partnership with India.
- 31. My Government looks forward to an enhanced partnership with India. FALSE
- 32. And more...
- 33. My Government will be reduced.
- 34. My Government will be strengthened in July.
- 35. My Government will be strengthened in Afghanistan.
- 36. People will be reviewed.
- 37. My Government will introduce fixed term Parliaments of unnecessary laws.
- 38. My Government will introduce fixed term Parliaments of children.
- 39. Legislation will be introduced for a viable Palestinian state existing in Northern Ireland.
- 40. My Government will support the abolition of Wales.
- 41. My Government looks forward to the blessing of Identity Cards and political parties.
- 42. The Almighty God may rest upon the management of the Commons.
- 43. Perception: real but implausible Can we analyse empathy in language? Can we analyse emotional reactions to speeches? Do speeches by MPs elicit the same reaction? A language uncanny valley?
- 44. Hubris Syndrome? P Garrard, D Owen, et al., “Linguistic Biomarkers of Hubris Syndrome, Cortex, Elsevier, 2013 COPYRIGHT BY THE ARTICLE’ S AUTHORS SEE FOOTNOTE FOR DETAILS
- 45. Stylistic similarity in literature J Hughes et al., “Quantitative patterns of stylistic influence in the evolution of literature”, PNAS, 2012 COPYRIGHT BY THE ARTICLE’ S AUTHORS SEE FOOTNOTE FOR DETAILS
- 46. Vocabulary Size M Daniels, “The largest vocabulary in Hip-Hop”, http://rappers.mdaniels.com.s3- website-us-east-1.amazonaws.com/ COPYRIGHT BY THE ARTICLE’S AUTHORS SEE FOOTNOTE FOR DETAILS
- 47. Reading level Fox et al., “Who was America’s most well-spoken President?” http://www.vocativ. com/interactive/usa/us-politics/presidential-readability/ COPYRIGHT BY THE ARTICLE’S AUTHORS SEE FOOTNOTE FOR DETAILS
- 48. Reading level Fox et al., “Who was America’s most well-spoken President?” http://www.vocativ. com/interactive/usa/us-politics/presidential-readability/ COPYRIGHT BY THE ARTICLE’S AUTHORS SEE FOOTNOTE FOR DETAILS
- 49. What lies ahead
- 50. Vocabulary Size
- 51. Ongoing Work Run away from PHP Python + NLTK Make it stable; auto-harvest Search on N-Grams, N>1 Search by MP and political party -> build each MP’s political history Build a base of analyses UI for the public + researchers