1. ML & Social Media
4th Summer School on Machine Learning
July 13, 2019
Ponnurangam Kumaraguru (“PK”)
IIIT Delhi
TEDx & ACM Distinguished Speaker
Linkedin/in/ponguru/
fb/ponnurangam.kumaraguru
@ponguru

2. 2

3. Keywords you got to know in last 120hrs?
3

4. 4

5. Top sites Alexa…
5

6. 6

7. 7

8. 8

9. 9
2009

10. 10

11. 11

12. 12

13. Non-trustworthy Content
FAKE
13
$
RUMORS

14. Methodology
14

15. Credibility Modeling
15
Feature set Features (45)
Tweet meta-data
Number of seconds since the tweet; Source of tweet (mobile / web/
etc); Tweet contains geo-coordinates
Tweet content (simple)
Number of characters; Number of words; Number of URLs; Number of
hashtags; Number of unique characters; Presence of stock symbol;
Presence of happy smiley; Presence of sad smiley; Tweet contains
`via'; Presence of colon symbol
Tweet content (linguistic)
Presence of swear words; Presence of negative emotion words;
Presence of positive emotion words; Presence of pronouns; Mention
of self words in tweet (I; my; mine)
Tweet author Number of followers; friends; time since the user if on Twitter; etc.
Tweet network
Number of retweets; Number of mentions; Tweet is a reply; Tweet is a
retweet
Tweet links WOT score for the URL; Ratio of likes / dislikes for a YouTube video

16. Implementation

17. Feedback by Users
17

18. Harvard (1839) – Harvard – Harvard – Harvard – MIT –
Northwestern – UIUC – WUSL – CMU (2009) – IIITD
(2015)
18

19. Different methods
– #Hashtag level
– Tweet level
– SVM
– Decision tree
– Sentiment analysis
– Text CNN
– LSTM
– Autoencoders
– Language models
– Knowledge Graph
– Multimodal
– GROVER
19
https://grover.allenai.org/

20. Different in Facebook…
20

21. Facebook Inspector
21

22. Harvard (1839) – Harvard – Harvard – Harvard – MIT –
Northwestern – UIUC – WUSL – CMU (2009) – IIITD
(Nov 14, 2017)
22

23. Harvard (1839) – Harvard – Harvard – Harvard – MIT –
Northwestern – UIUC – WUSL – CMU (2009) – IIITD
(April 25, 2019)
23

24. Pittsburgh Shooting / Gab post
24

25. Hate speech
25

26. Code mixed content
26

27. Code mixed content
27

28. Stance prediction
28

29. De-duplicating audience
Social audience = 437,632 + 153,000 + 805,097 or less??
29

30. 30

31. 31

32. Heuristic Identity Search
32
cerc.iiitd.ac.in
Profile
Content
Self-mention
Network
Syntactic
and Image
Search Linking
If self-identified /
returned by
more than one
search method
No
Yes
Candidate
Identities
name,
location,
username
mobile no,
post,
friends,
followers
Paridhi Jain, Ponnurangam Kumaraguru, and Anupam Joshi. 2013. @I seek ‘fb.me’: Identifying Users across Multiple Online Social
Networks. In Proceedings of the 22nd International Conference on World Wide Web, WWW ’13 Companion. ACM, New York, NY, USA,
1259- 1268. DOI=http://dx.doi.org/10.1145/2487788.2488160 [Honorable Mention Award}

33. Harvard (1839) – Harvard – Harvard – Harvard – MIT –
Northwestern – UIUC – WUSL – CMU (2009) – IIITD
(2016)
33

34. WTF / Recommendations….
34

35. YouTube Recommendations….
35

36. To understand users / customers
– “Show me who your friends are and I’ll tell
you who you are?”
36

37. To understand users / customers
– “Show me who your friends are and I’ll tell
you who you are?”
37

38. To understand users / customers
– “Show me who your friends are and I’ll tell
you who you are?”
38

39. To understand users / customers
– Interest graph
39

40. 40
Selfies vs Potential KillFies

41. 41
Selfies vs Potential KillFies

42. 42

43. 439th International Workshop on Modeling Social Media (MSM), 2018 @ WWW

44. 44

45. 45
http://bit.ly/saftie-cam

46. Sample
46

47. Takeaways
– Social Media is a powerful tool
– Business / Government can use the analysis
to their benefits
– You should use it for your benefits J
-You tried doing analysis on your accounts?
– Lots of challenges and opportunities!
– Will be happy to interact more, if needed
47

48. 48
pk@iiitd.ac.in
http://precog.iiitd.edu.in/
https://www.instagram.com/pkatiiitd/
https://fb.com/ponnurangam.kumaraguru
@ponguru
Linkedin/in/ponguru/