2. Zipf 's Law
Given some corpus of natural language utterances, the frequency (f), of
any word is inversely proportional to its rank (R), in the frequency table.
R
f
1
3. Zipf 's Law for different English authors
(From work of G. K. Zipf, 1949)
4. Zipf 's Law in Multilingual Wikipedia
(From work of Sergio Jimenez, 2015)
5. Zipf 's Law in Science and Facebook
(From work of Néda et al, 2017)
f(x) is the probability of a paper(or Facebook page) to have x citations(or shares).
6. Zipf 's Law in Messenger Chat
0,0 0,5 1,0 1,5 2,0
0,0
0,5
1,0
1,5
2,0
2,5
3,0
Different friends
Linear fit
Log(Numberofmessages)
Log (Rank of friend)
A better modelling is possible using the Tsallis-Pareto distribution.
(Newman, 2005)
8. Applications : Managing Depression
and Social Health
People who were classified as at risk for depression had less rewarding
interactions than people who were not at risk (Nezlek, 1994).
Positive social interactions can exert substantial benefits on health.
(Aguilar-Raab, 2018).
Patients with small social networks had an elevated risk of mortality.
(Brumett, 2001).
9. Applications : Social Health Evaluation
(From World Health Organization, 2017)
• Study chat history for people with depression.
• Calibrate a model for social health(S H) index with respect to
deviation from Zipf's law.
• Monitor S H index for users.
• Notify the user when S H index reaches a critical limit.
• Cater increased socializing options to manage depression.
10. S C A : Socializing Options
Automated Ping
• S C A will monitor user's chat statistics.
• It will calculate rank of all of the user's friends.
• It will check when the last conversation with a friend ended.
• Based on the rank of the friend and the Zipf's law model, if a
friend is not in touch more than a certain time, it will send a
«hi» or «how are you doing» message to avoid social
seclusion.
11. S C A : Socializing Options
Personalized Automated Ping
My friend Dr. Shonku
Ping frequency (pf) for Dr. Shonku
(linear model)
pf = A0Z0 + A1Z1 + A2Z2 + ...
Z0 is the calculated ping frequency based on the Zipf's law model. A0 is set to 1 by
default. Zn (n>0) are other parameters like 'closeness', 'imortance', etc. An (n>0) can
be set by the user as a value between -1 and 1.
12. S C A : Socializing Options
Personalized Automated Ping : Demo
13. S C A : Socializing Options
Digital Friend
(Developed by Rollo Carpenter)
14. S C A : Socializing Options
Digital Friend (sample conversation)
15. S C A : Socializing Options
Personalized Digital Friendship (in progress)
• S C A will collect user's chat data.
• Train a personalized neural network like Cleverbot.
• The chatbot will ping the user under abnormal social
behavior scenarios.
• It can be thought of as a next generation personal diary
with whom the user can engage and share feelings.
17. Conclusion
• S C A will keep track of a user's social health index.
• S C A can cater social opprotunities to tackle depression
and other mental disorders in a personalized way.
• S C A can train a neural network like Cleverbot to
simulate digital friends.
• Simulated digital friends can provide a personalized
interaction experience to the user and can make life
unique and awesome!