Structure of media attention
0 likes785 views
This document summarizes research on analyzing the structure of media attention networks using newspaper articles. Named entities are extracted from articles and linked based on co-occurrence, building a network that can reveal patterns of elite behavior and interactions. The network exhibits core-periphery structure, with high-degree "hubs" attracting more weight and connections. This structure likely arises from both the underlying bipartite network of articles and entities, as well as repetitive interactions between elites. Further analysis of network dynamics over time and across contexts could provide insight into how elite networks function.
Structure of media attention
- 1. The structure of media attention V.A. Traag, R. Reinanda, J. Hicks, G. Van Klinken KITLV, Leiden, the Netherlands e-Humanities, KNAW, Amsterdam, the Netherlands September 30, 2014 e Humanities Royal Netherlands Academy of Arts and Sciences
- 2. Background Research focus Study elite (network) behaviour. Relation with political developments. Data: newspaper articles. How can we use them? Data Current corpus: Joyo/Indonesian News Service, 2004{2012. Contains about 140 263 articles.
- 3. Network Building the network 1 Detect names automatically . I Budhisantoso would ask Kalla to team up with Yudhoyono . 2 Disambiguate names. I Susilo Bambang Yudhoyono or Dr. Yudhoyono , etc. . . 3 Co-occurrence in sentence (record frequency). I Budhisantoso would ask Kalla to team up with Yudhoyono . K 1 B Y 1 1
- 4. Network Building the network 1 Detect names automatically . I Budhisantoso would ask Kalla to team up with Yudhoyono . 2 Disambiguate names. I Susilo Bambang Yudhoyono or Dr. Yudhoyono , etc. . . 3 Co-occurrence in sentence (record frequency). I Budhisantoso would ask Kalla to team up with Yudhoyono . K 1 B Y 1 1
- 5. Network Building the network 1 Detect names automatically . I Budhisantoso would ask Kalla to team up with Yudhoyono . 2 Disambiguate names. I Susilo Bambang Yudhoyono or Dr. Yudhoyono , etc. . . 3 Co-occurrence in sentence (record frequency). I Budhisantoso would ask Kalla to team up with Yudhoyono . K 1 B Y 1 1
- 6. Network Building the network 1 Detect names automatically . I Budhisantoso would ask Kalla to team up with Yudhoyono . 2 Disambiguate names. I Susilo Bambang Yudhoyono or Dr. Yudhoyono , etc. . . 3 Co-occurrence in sentence (record frequency). I Budhisantoso would ask Kalla to team up with Yudhoyono . K 1 B Y 1 1
- 7. Strength 101 100 100 101 102 103 Degree Average weight Joyo NYT 100 101 102 103 104 Degree Data Hubs co-occur more frequently.
- 8. Strength 101 100 100 101 102 103 Degree Average weight Joyo NYT 100 101 102 103 104 Degree Data Bipartite Hubs co-occur more frequently.
- 9. Clustering 100 10−1 10−2 100 101 102 103 10−3 Degree Clustering Joyo NYT 100 101 102 103 104 Degree Data Hubs tend to cluster less.
- 10. Clustering 100 10−1 10−2 100 101 102 103 10−3 Degree Clustering Joyo NYT 100 101 102 103 104 Degree Data Bipartite Hubs tend to cluster less.
- 11. Clustering 100 10−1 100 101 102 103 Degree Weighted Clustering Joyo NYT 100 101 102 103 104 Degree Data Hubs tend to cluster less (also weighted).
- 12. Clustering 100 10−1 100 101 102 103 Degree Weighted Clustering Joyo NYT 100 101 102 103 104 Degree Data Bipartite Hubs tend to cluster less (also weighted).
- 13. Neighbour degree 103 102 100 101 102 103 101 Degree Neighbour Degree Joyo NYT 100 101 102 103 104 Degree Data Hubs tend to connect to low degree nodes.
- 14. Neighbour degree 103 102 100 101 102 103 101 Degree Neighbour Degree Joyo NYT 100 101 102 103 104 Degree Data Bipartite Hubs tend to connect to low degree nodes.
- 15. Weighted Neighbour degree 103 102 100 101 102 103 Degree Weighted Neighbour Degree Joyo NYT 100 101 102 103 104 Degree Data But hubs connect much stronger to other hubs.
- 16. Weighted Neighbour degree 103 102 100 101 102 103 Degree Weighted Neighbour Degree Joyo NYT 100 101 102 103 104 Degree Data Bipartite But hubs connect much stronger to other hubs.
- 17. Predict weight 100 101 102 103 104 104 103 102 101 100 Weight Predicted Weight Joyo NYT 100 101 102 103 104 Weight Data wij J ij exp((si sj )
- 18. )
- 19. Predict weight 100 101 102 103 104 104 103 102 101 100 Weight Predicted Weight Joyo NYT 100 101 102 103 104 Weight Data Bipartite wij J ij exp((si sj )
- 20. )
- 21. Core-periphery Summary Results Hubs attract much more weight. Most of the weight between hubs. Low degree node connect to hubs. Low degree nodes cluster locally. Consistent with core-periphery structure. But, seems also present in bipartite randomisation. Largest deviations, empirically: Degree is lower, average weight is higher. Weighted neighbour degree increases.
- 22. Model Simple model to overcome deviations: 1 Create empty sentence 2 Add certain number of nodes 1 Either random node (with PA) 2 Or random neighbour (with PA) Probability (ki + 1)
- 23. . 3 Repeat
- 24. Degree Weight Empirical Bipartite Model Joyo Avg. Degree 12:4 22:1 12:2 Avg. Weight 2:9 1:2 2:8 NYT Avg. Degree 22:3 45:2 22:6 Avg. Weight 2:01 1:11 1:31
- 25. Strength 101 100 100 101 102 103 Degree Average weight Joyo NYT 100 101 102 103 104 Degree Data Model Weight increases more in the model.
- 26. Weighted neigbhour degree 103 102 100 101 102 103 Degree Weighted Neighbour Degree Joyo NYT 100 101 102 103 104 Degree Data Bipartite Weighted neighbour degree increases in the model.
- 27. Conclusions Results: Network looks like core-periphery. Probably due to bipartite structure. But also to repetitive interaction. Further research: Basis for comparing elite networks. Compare networks across time and space. Dynamical, temporal aspects.
- 28. Thank you! Questions? Presentation: SlideShare Paper: arXiv:1409.1744 Dynamics of network: arXiv:1409.2973 http://www.traag.net @vtraag