Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph...
Transcript of Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph...
![Page 1: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/1.jpg)
Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks
René Pickhardt, Thomas Gottron, Jonas Kunze, Ansgar Scherp Steffen Staab
How to retrieve more than 10'000 temporal ordered news feeds per second in social networks with millions of users like Facebook and Twitter by using graph
data bases (like neo4j) and Graphity
![Page 2: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/2.jpg)
Joint collaboration with
Thomas Gottron
JonasKunze (from metalcon.de)
Steffen Staab
Ansgar Scherp
![Page 3: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/3.jpg)
Index
• Introduction to the newsfeed problem
• Why relational Data bases won't do the job
• The construction and idea of STOU
• The construction and idea of graphity
• Example 1: retrieval of news feeds (top-k n-way merge)
• Example 2: Creating new Content Items
• Evaluation on Wikipedia data set.
![Page 4: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/4.jpg)
A "typical" social network graph
![Page 5: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/5.jpg)
Retrieving Node A's news stream
![Page 6: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/6.jpg)
Retrieving Node A's news stream
![Page 7: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/7.jpg)
Some Challenges
Social networks like Twitter and Facebook have several thousand requested news feeds per second
News feeds change fast: Several hundred newly created content items per second. (600 tweets / sec in 2010)
News feeds are different for every user
Realtime (retrieval should be as low as micro seconds)
Friendship graph changes over time
Overall: This is a very dynamic problem with a lot of chaotic & unpredictable behaviour
![Page 8: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/8.jpg)
Index
• Introduction to the newsfeed problem
• Why relational Data bases won't do the job
• The construction and idea of STOU
• The construction and idea of graphity
• Example 1: retrieval of news feeds (top-k n-way merge)
• Example 2: Creating new Content Items
• Evaluation on Wikipedia data set.
![Page 9: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/9.jpg)
First we have some Users in a social Network
ID
abcde
User
![Page 10: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/10.jpg)
They follow other users
ID
abcde
from toa ca ba db cb db e
User Follower
![Page 11: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/11.jpg)
Everyone produces status updates and content
ID
abcde
from toa ca ba db cb db e
User time Content
d 19 Lorem ipsum
e 18 dolor sit amet,
e 17 consectetur
b 14 adipisici elit, sed
a 13 eiusmod tempor
c 12 incidunt ut labore
b 11 et dolore magna
a 8 aliqua. Ut enim
c 5 ad minim veniam
d 3 quis nostrud
User Follower
ContentItems
![Page 12: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/12.jpg)
Our Query joins over huge Follower Matrix
ID
abcde
from toa ca ba db cb db e
User time Content
d 19 Lorem ipsum
e 18 dolor sit amet,
e 17 consectetur
b 14 adipisici elit, sed
a 13 eiusmod tempor
c 12 incidunt ut labore
b 11 et dolore magna
a 8 aliqua. Ut enim
c 5 ad minim veniam
d 3 quis nostrud
User Follower
ContentItem
SELECT ci.User, ci.time, ci.Content FROM ContentItems ci JOIN Follower f on ci.User=f.to JOIN User u on u.ID = f.from WHERE u.ID like "a" ORDER BY ci.time DESC
![Page 13: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/13.jpg)
Index
• Introduction to the newsfeed problem
• Why relational Data bases won't do the job
• The construction and idea of STOU
• The construction and idea of graphity
• Example 1: retrieval of news feeds (top-k n-way merge)
• Example 2: Creating new Content Items
• Evaluation on Wikipedia data set.
![Page 14: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/14.jpg)
STOU = Star Topology (Temporal) Ordered Updates
From the standard social network graph we move to (temporal) ordered lists for rectangular nodes
![Page 15: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/15.jpg)
Introducing linked lists for status updates!
What are the pros and cons of this change?
![Page 16: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/16.jpg)
• dynamic retrival possible (friendship graph may change)• very flexible data structure
• inserts and removes are very fast (all operations are O(1))
Pros of this approach
![Page 17: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/17.jpg)
• unclear which edge to traverse first!• ==> entire ego network must be sorted
• Size d of an ego network is usually much bigger than the number of retrieved items k.
• ==> Sorting seems to much effort
Contra: slow retrieval --> O(d log(d))
![Page 18: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/18.jpg)
Index
• Introduction to the newsfeed problem
• Why relational Data bases won't do the job
• The construction and idea of STOU
• The construction and idea of graphity
• Example 1: retrieval of news feeds (top-k n-way merge)
• Example 2: Creating new Content Items
• Evaluation on Wikipedia data set.
![Page 19: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/19.jpg)
The key concept: going from star topology to lists
![Page 20: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/20.jpg)
graphity index for the node "a"
Graphity rules:
• for every node (a & b) that follows others we create a linked list
• this linked list contains all the nodes that are beeing followed by this node.
• The followed nodes are sorted by the timestamp of their most recent content item
![Page 21: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/21.jpg)
second graphity index for node "b"
![Page 22: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/22.jpg)
Index
• Introduction to the newsfeed problem
• Why relational Data bases won't do job
• The construction and idea of STOU
• The construction and idea of graphity
• Example 1: retrieval of news feeds (top-k n-way merge)
• Example 2: Creating new Content Items
• Evaluation on Wikipedia data set.
![Page 23: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/23.jpg)
Rearanging the graph ...
![Page 24: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/24.jpg)
top-k n-way Merge for retrieval in O(k log(k))
push(19) to PQ
Pointer tmp
Stream:
PriorityQueue:
19
![Page 25: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/25.jpg)
top-k n-way Merge for retrieval in O(k log(k))
Pointer tmp
push(14) to PQ
push(3) to PQ
Stream: (19,d)
PriorityQueue:
143
![Page 26: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/26.jpg)
top-k n-way Merge for retrieval in O(k log(k))
Pointer tmp
push(11) to PQ
Stream: (19,d) (14,b)
PriorityQueue:
12113
push(12) to PQ
![Page 27: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/27.jpg)
top-k n-way Merge for retrieval in O(k log(k))
Pointer tmp
push(5) to PQ
Stream: (19,d) ; (14,b) ; (12,c)
PriorityQueue:
1153
![Page 28: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/28.jpg)
top-k n-way Merge for retrieval in O(k log(k))
Stream: (19,d) ; (14,b) ; (12,c) ; ...
PriorityQueue:
1153
retrieve 11 now!
![Page 29: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/29.jpg)
Index
• Introduction to the newsfeed problem
• Why relational Data bases won't do the job
• The construction and idea of STOU
• The construction and idea of graphity
• Example 1: retrieval of news feeds (top-k n-way merge)
• Example 2: Creating new Content Items
• Evaluation on Wikipedia data set.
![Page 30: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/30.jpg)
Creating new Content items
b creates a new content item
![Page 31: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/31.jpg)
Creating new Content items
b created 20
• update linked list of b's content items
![Page 32: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/32.jpg)
Creating new Content items
b created 20
• update linked list of b's content items
• now look in which ego networks b is member of. (our case just a)
![Page 33: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/33.jpg)
Creating new Content items
b created 20
• update linked list of b's content items
• now look in which ego networks b is member of. (our case just a)
• interlink b's predecessor and successor
![Page 34: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/34.jpg)
Creating new Content items
b created 20
• update linked list of b's content items
• now look in which ego networks b is member of. (our case just a)
• interlink b's predecessor and successor
• user the follow edge from a to b and the first ego:a to insert b in the beginning of ego:a
![Page 35: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/35.jpg)
Index
• Introduction to the newsfeed problem
• Why relational Data bases won't do the job
• The construction and idea of STOU
• The construction and idea of graphity
• Example 1: retrieval of news feeds (top-k n-way merge)
• Example 2: Creating new Content Items
• Evaluation on Wikipedia data set.
![Page 36: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/36.jpg)
Wikipedia as a social network
• Every article ==> User
• Every link in an article ==> Follow relationship
• Every Revision of an article ==> Status update of a user
• Remark: if in a new revision the outlinks of the wikipedia article change
• We don't take this as a status update• we interpret this revision as a change to the friendship
graph
![Page 37: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/37.jpg)
demonstrating independence of node degree
![Page 38: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/38.jpg)
demonstrating linear dependence of k
![Page 39: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/39.jpg)
Index maintaining - inserting new content items
![Page 40: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/40.jpg)
updating graphity for new friendship relations
![Page 41: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/41.jpg)
updating graphity if friendships break
![Page 42: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/42.jpg)
time to build the index
![Page 43: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/43.jpg)
Conclusion
• We built two graph indices for top-k news feed retrieval
• STOU is fastest in writing operations (with moderate reading speed)
• Graphity is faster in retrieving operations
• Empirical study on a graphs with up to 2 mio. Users, 32 mio. follow relationships and 50 mio. content items shows that graphity even performs better than the theoretical runtime
• Especially for graphity we saw• retrieval of social news feeds of k items in O(k log(k))• Almost as good as redundant content lists• But no redundancy in content data
![Page 44: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/44.jpg)
So which one to take?
42
![Page 45: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/45.jpg)
Thank you for your attention
More information + Slides on:http://www.rene-pickhardt.de/graphity
Thanks to • Mattias Persson and Peter Neubauer from neotechnology.com • the neo4j community on the neo4j mailinglist for helpful advices • Knut Schumach for coming up with the name GRAPHITY • Matthias Thimm & Leon Kastler for helpful discussions
This project is founded by the EU Projects Social Sensor and ROBUST.
Sorce code & data sets on:http://www.rene-pickhardt.de/graphity-source-code/
![Page 46: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/46.jpg)
Backup slides
• Backup slides
![Page 47: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/47.jpg)
Rearanging the graph...
![Page 48: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/48.jpg)
Updates need to be done in the following situation
Updates need to be done in the following situations
• new created content item (O(d))o index of every follower needs to be updated
• new created follow relation (O(d))o index of follower needs to be updated
• friendship relation breaks (O(d))o index of the former follower needs to be updated
• most recent content item of a user is deleted (O(d²))o index of every follower nees to be updated
![Page 49: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/49.jpg)
Future work
• Generalize / built theory on top-k joins
• Distributed system
• Partially do graphity index • (ever update only yields updating a constant number
of graphity indices)
• Tie strength (filtering / ranking)
![Page 50: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/50.jpg)
Evaluation Characteristics of data sets
![Page 51: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/51.jpg)
Retrieving streams on all data sets
![Page 52: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/52.jpg)
Retrieving streams Node degre > 10
![Page 53: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/53.jpg)
Simulating a social network
![Page 54: Efficient Graph Models for Retrieving the Top- k News Feeds … · 2018-08-15 · Efficient Graph Models for Retrieving the Top-k News Feeds from Ego Networks René Pickhardt, Thomas](https://reader034.fdocuments.us/reader034/viewer/2022042310/5ed74a54c079a632805800a7/html5/thumbnails/54.jpg)
Summary - We created a graph model with:
• fast retrieval of social news feeds of k items in O(k log(k))• dynamic retrieval method• no redundancy in content data• Creating new Status Updates yields updating of d graphity
indices of following nodes• Each Graphity index update is O(1)
We also conducted an evaluation of a graph with :• ~ 2 mio. users• ~32 mio. follow relations• ~50 mio. Status updates
giving empirical proof of our theoretical findings.