1. Live Social Semantics
& online community monitoring
Harith Alani
Knowledge Media institute,
The Open University, UK
http://twitter.com/halani
http://delicious.com/halani
http://www.linkedin.com/pub/harith-alani/9/739/534
Semantic Web Summer School
Cercedilla, Spain,, 2011 1
3. Location, Sensors, & Social Networking
Tag-Along Marketing
The New York Times,
November 6, 2010
“Everything is in place for location-based social
networking to be the next big thing. Tech
companies are building the platforms, venture
capitalists are providing the cash and marketers
are eager to develop advertising. “
3
4. Location, Sensors, & Social Networking
The Canine Twitterer
“Having my daily
workout. Already did
15 leg lifts!”
4
10. Online vs. offline social networking
• Digital social networking • Digital networking increase
increases physical social social interaction
isolation – Create more opportunities to
network
• Causes – Supports and increases F2F
– Genetic alterations contact!
– Weakened immune system – Stronger offline social tiesà
– Less resistant to cancer more online communication
– Higher risk of heart disease – Stronger offline social ties à
– Higher blood pressure more diverse online
– Faster dementia communications
– Narrower arteries – F2F is medium of choice in
weaker social ties
Barry Wellman, The Glocal Village: Internet and
Aric Sigman, “Well Connected? The Biological Community, Idea’s - The Arts & Science Review,
Implications of 'Social Networking’”, Biologist, 56 University of Toronto, 1(1),2004
(1), 2009
10
11. Offline + online social networking
Who should
Anyone I I talk to? Where have I
know here? met this guy?
Where
should I go?
ESWC2010 11
12. Live Social Semantics (LSS):
RFIDs + Social Web + Semantic Web
<?xml version="1.0"?>!
<rdf:RDF!
xmlns="http://
tagora.ecs.soton.ac.uk/schemas/
tagging#"!
xmlns:rdf="http://www.w3.org/
1999/02/22-rdf-syntax-ns#"!
xmlns:xsd="http://www.w3.org/2001/
XMLSchema#"!
xmlns:rdfs="http://www.w3.org/
2000/01/rdf-schema#"!
xmlns:owl="http://www.w3.org/
2002/07/owl#"!
xml:base="http://
tagora.ecs.soton.ac.uk/schemas/
tagging">!
<owl:Ontology rdf:about=""/>!
<owl:Class rdf:ID="Post"/>!
<owl:Class rdf:ID="TagInfo"/>!
<owl:Class
rdf:ID="GlobalCooccurrenceInfo"/>!
<owl:Class
rdf:ID="DomainCooccurrenceInfo"/>!
<owl:Class rdf:ID="UserTag"/>!
<owl:Class
rdf:ID="UserCooccurrenceInfo"/>!
<owl:Class rdf:ID="Resource"/>!
<owl:Class rdf:ID="GlobalTag"/>!
<owl:Class rdf:ID="Tagger"/>!
<owl:Class rdf:ID="DomainTag"/>!
<owl:ObjectProperty
rdf:ID="hasPostTag">!
<rdfs:domain
rdf:resource="#TagInfo"/>!
</owl:ObjectProperty>!
<owl:ObjectProperty
rdf:ID="hasDomainTag">!
<rdfs:domain
rdf:resource="#UserTag"/>!
</owl:ObjectProperty>!
<owl:ObjectProperty
rdf:ID="isFilteredTo">!
• Integration of physical presence and online information
<rdfs:range
rdf:resource="#GlobalTag"/>!
<rdfs:domain
• Semantic user profile generation
rdf:resource="#GlobalTag"/>!
</owl:ObjectProperty>!
<owl:ObjectProperty
• Logging of face-to-face contactrdf:ID="hasResource">!
<rdfs:domain rdf:resource="#Post"/>!
<rdfs:range =…!
• Social network browsing
• Analysis of online vs offline social networks
13. m Live Social Semantics: architecture Communities of Practice
Communities of Practice
dbtune.org rkbexplorer.com
Publications Profile Builder dbpedia.org Publications Profile
org semanticweb.org
ontology
Web-based Systems
Profile interests
data.semanticweb.org TAGora Sense
builderDelicious
rkbexplorer.com Repository
Extractor Extractor
publications, co-authorship networks Flickr
Daemon Social Tagging mbid -> dbpedia uri Daemon Social Tagging
Social Networks tag -> dbpedia uri Social Networks
LastFM
Connect API JXT Triple Store Facebook Connect API JXT Trip
Contacts social semantics Contacts
URIs
Tag disambiguation
Social triple store Social
service Semantics
RDF cache
Aggregator
Semantics
RDF cache
contacts data
RFID
Local Local
Readers
Real World
Server Server Real-World
Tag to URI Real-World
networks
service
Contact Data Contact Data
tags
RFID
Badges
Visualization Web Interface Linked Data Visualization Web In
Linked data Web interface Visualization
13
27. Live Social Semantics
Deployed at:
Data analysis
• Face-to-face interactions across scientific conferences
• Networking behaviour of frequent users
• Correlations between scientific seniority and social networking
• Comparison of F2F contact network with Twitter and Facebook
• Social networking with online and offline friends
27
28. Characteristics of F2F contact network
Network ESWC 2009 HT 2009 ESWC 2010
characteristics
Number of users 175 113 158
Average degree 54 39 55
Avg. strength (mn) 143 123 130
Avg. weight (mn) 2.65 3.15 2.35
Weights ≤ 1 mn 70% 67% 74%
Weights ≤ 5 mn 90% 89% 93%
Weights ≤ 10 mn 95% 94% 96%
• Degree is number of people with whom the person had at least one F2F
contact
• Strength is the time spent in a F2F contact
• Edge weight is total time spent by a pair of users in F2F contact
28
29. Characteristics of F2F contact events
Contact ESWC 2009 HT 2009 ESWC 2010
characteristics
Number of 16258 9875 14671
contact events
Average contact 46 42 42
length (s)
Contacts ≤ 1mn 87% 89% 88%
Contacts ≤ 2mn 94% 96% 95%
Contacts ≤ 5mn 99% 99% 99%
Contacts ≤ 10mn 99.8% 99.8% 99.8%
F2F contact pattern is very similar for all three conferences
30. F2F contacts of returning users
Degree
• Degree: number of other 10
2
participants with whom an attendee
has interacted
1
10 1 2
10 10
• Total time: total time spent in
ESWC2010
Total interaction time
interaction by an attendee 4
10
3
10 3 4 5
10 10 10
• Link weight: total time spent in F2F 4 Links’ weights
10
interaction by a pair of returning 3
10
attendees in 2010, versus the same 2
10
quantity measured in 2009 1
10 1 2 3 4 5
10 10 10 10 10
ESWC 2009 & Pearson Correlation ESWC2009
ESWC 2010
Degree 0.37 Time spent on F2F networking by frequent
users is stable, even when the list of
Total F2F 0.76
interaction time people they networked with changed
Link weight 0.75
30
31. Average seniority of neighbours in F2F networks
• No clear pattern is observed 5
if the unweighted average senn
Avg seniority of the neighbours
over all neighbours in the
Average seniority of neighbors
senn,w
with weighted averages
aggregated network is 4
considered
senn,max
Seniority of user with strongest link
• A correlation is observed 3
when each neighbour is
weighted by the time spent
with the main person
2
• The correlation becomes
much stronger when 1
considering for each
individual only the neighbour
with whom the most time was
spent 0
0 5 10
seniority (number of papers)
Conference attendees tend to networks with others of similar
levels of scientific seniority
31
33. Offline networking vs online networking
Twitterers Spearman
Correlation (ρ)
Tweets – F2F Degree - 0.15
Tweets – F2F Strength - 0.15
Twitter Following – F2F - 0.21
Degree
users
Users with Facebook and Twitter accounts in ESWC 2010
• people who have a large number of friends on Twitter and/or Facebook don’t seem to
be the most socially active in the offline world in comparison to other SNS users
No strong correlation between amount of F2F
contact activity and size of online social networks 33
34. Scientific seniority vs Twitter followers
Twitter users Correlation
H-index – Twitter Followers 0.32
(#$"
H-index – Tweets - 0.13
("
!#'"
*+,-./"01221+./3"
!#&"
45678.9"
*+..:3"
!#%"
!#$"
!"
(" &" ((" (&" $(" $&" )(" )&" %(" users
• Comparison between people’s scientific seniority and the number of people following
them on Twitter
People who have the highest number of Twitter followers are not
necessarily the most scientifically senior, although they do have high
visibility and experience 34
35. Conference Chairs
all chairs all chairs
participants 2009 participants 2010
2009 2010
average degree 55 77.7 54 77.6
average strength 8590 19590 7807 22520
average weight 159 500 141 674
average number of 3.44 8 3.37 12
events per edge
• Conf chairs interact with more distinct people (larger average degree)
• Conf chairs spend more time in F2F interaction (almost three times as much
as a random participant)
36. Networking with online and offline ‘friends’
Characteristics all users coauthors Facebook Twitter
friends followers
average contact 42 75 63 72
duration (s)
average edge weight 141 4470 830 1010
(s)
average number of 3.37 60 13 14
events per edge
• Individuals sharing an online or professional social link meet much more
often than other individuals
• Average number of encounters, and total time spent in interaction, is highest
for co-authors
F2F contacts with Facebook & Twitter friends were respectively %50 and
%71 longer, and %286 and %315 more frequent than with others
They spent %79 more time in F2F contacts with their co-authors, and they
met them %1680 more times than they met non co-authors
37. Twitterers vs Non-Twitterers
• Time spent in conference rooms
– Twitter users spent on average 11.4% more time in the
conf rooms than non-twitter users (mean is 26% higher)
• Number of people met F2F during the conference
– Twitter users met on average 9% more people F2F
(mean 8% higher)
• Duration of F2F contacts
– Twitter users spent on average 63% more time in F2F
contact than non twitter users (mean is 20% higher)
37
39. Behaviour analysis
Jeffrey Chan, Conor Hayes, and Elizabeth Daly. Decomposing discussion forums using
common user roles. In Proc. Web Science Conf. (WebSci10), Raleigh, NC: US, 2010
42. Approach for inferring User Roles
Structural, social network, Feature levels change with the
reciprocity, persistence, participation dynamics of the community
Run our rules over each user’s features Associate Roles with a collection of
and derive the role composition feature-to-level Mappings
e.g. in-degree -> high, out-degree ->
high
42
43. Data from Boards.ie
• Forum 246 (Commuting and Transport): Demonstrates a clear increase in
activity over time.
• Forum 388 (Rugby): Exhibits periodic increase and decrease in activity and
hence it provides good examples of healthy/unhealthy evolutions.
• Forum 411 (Mobile Phones and PDAs): Increase in activity over time with
some fluctuation - i.e. reduction and increase over various time windows.
• For the time in 2004-01 to 2006-12
44. Results
Commuting and Transport Rugby Mobile Phones and PDAs
• Correlation of individual features in each of the three forums
45. (a) Forum 246: Commuting and Transport
Results
(b) Forum 388: Rugby
(c) Forum 411: Mobile Phones and PDAs
• Variation in behaviour
composition & activity
• Behaviour composition in/
stability influences forum
activity
46. Prediction analysis – preliminary results!
• Predicting rise/fall in post submission numbers
• Binary classification
• Features : Community composition, roles and percentages of users
associated with each
Forum P R F1 ROC
246 0.799 0.769 0.780 0.800
388 0.603 0.615 0.605 0.775
411 0.765 0.692 0.714 0.617
All 0.583 0.667 0.607 0.466
• Cross-community predictions are less reliable than individual
community analysis due to the idiosyncratic behaviour observed in
each individual community
48. Predicting engagement
• Which posts will receive a reply?
– What are the most influential features here?
• How much discussion will it generate?
– What are the key factors of lengthy discussions?
48
49. user attributes - describing the reputation of the user - and attributes of a post’s
content - generally referred to as content features. In Table 1 we define user and
Common online communityFeatures features
content features and study their influence on the discussion “continuation”.
Table 1. User and Content
User Features
In Degree: Number of followers of U #
Out Degree: Number of users U follows #
List Degree: Number of lists U appears on. Lists group users by topic #
Post Count: Total number of posts the user has ever posted #
User Age: Number of minutes from user join date #
P ostCount
Post Rate: Posting frequency of the user U serAge
Content Features
Post length: Length of the post in characters #
Complexity: Cumulative entropy of the unique words in post p λ
i∈[1,n] pi(log λ−log pi)
of total word length n and pi the frequency of each word λ
Uppercase count: Number of uppercase words #
Readability: Gunning fog index using average sentence length (ASL) [7]
and the percentage of complex words (PCW). 0.4(ASL + P CW )
Verb Count: Number of verbs #
Noun Count: Number of nouns #
Adjective Count: Number of adjectives #
Referral Count: Number of @user #
Time in the day: Normalised time in the day measured in minutes #
Informativeness: Terminological novelty of the post wrt other posts
The cumulative tfIdf value of each term t in post p t∈p tf idf (t, p)
Polarity: Cumulation of polar term weights in p (using
P o+N e
Sentiwordnet3 lexicon) normalised by polar terms count |terms|
• How do all these features influence activity generation in an online
4.2 Experiments
community? are intended to test the performance of different classification mod-
Experiments
– els in identifying seed posts. Therefore we used four classifiers: discriminative
Such knowledge leads to better use and management of the community 49
classifiers Perceptron and SVM, the generative classifier Naive Bayes and the
50. Experiment for identifying Twitter seed posts
• Twitter data on the Haiti earthquake, and the Union
Address
Dataset Users Tweets Seeds Non-seeds Replies
Haiti 44,497 65,022 1,405 60,686 2,931
Union Address 66,300 80,272 7,228 55,169 17,875
• Evaluated a binary classification task
– Is this post a seed post or not?
50
51. first report on the results obtained from our model selection phase, before moving
Identifying seeds with different type of
onto our results from using the best model with the top-k features.
features
Table 3. Results from the classification of seed posts using varying feature sets and
classification models
(a) Haiti Dataset (b) Union Address Dataset
P R F1 ROC P R F1 ROC
User Perc 0.794 0.528 0.634 0.727 User Perc 0.658 0.697 0.677 0.673
SVM 0.843 0.159 0.267 0.566 SVM 0.510 0.946 0.663 0.512
NB 0.948 0.269 0.420 0.785 NB 0.844 0.086 0.157 0.707
J48 0.906 0.679 0.776 0.822 J48 0.851 0.722 0.782 0.830
Content Perc 0.875 0.077 0.142 0.606 Content Perc 0.467 0.698 0.560 0.457
SVM 0.552 0.727 0.627 0.589 SVM 0.650 0.589 0.618 0.638
NB 0.721 0.638 0.677 0.769 NB 0.762 0.212 0.332 0.649
J48 0.685 0.705 0.695 0.711 J48 0.740 0.533 0.619 0.736
All Perc 0.794 0.528 0.634 0.726 All Perc 0.630 0.762 0.690 0.672
SVM 0.483 0.996 0.651 0.502 SVM 0.499 0.990 0.664 0.506
NB 0.962 0.280 0.434 0.852 NB 0.874 0.212 0.341 0.737
J48 0.824 0.775 0.798 0.836 J48 0.890 0.810 0.848 0.877
4.3 Results
Our• findings from Table 3 demonstrate the effectiveness of using solely user
User features are most important in Twitter
features for identifying seed posts. Infeatures gives best results Address datasets
• But combining user & content both the Haiti and Union
training a classification model using user features shows improved performance51
over the same models trained using content features. In the case of the Union
52. Impact of different features in Twitter
which we found to be 0.674 indicating a good correlation between the two lists
and• their respective ranks.the highest impact on identification of seed
What features have
posts?
TableRank features by information gainGain Ratio wrt Seed Post class label. The
• 4. Features ranked by Information ratio wrt seed post class label
feature name is paired within its IG in brackets.
Rank Haiti Union Address
1 user-list-degree (0.275) user-list-degree (0.319)
2 user-in-degree (0.221) content-time-in-day (0.152)
3 content-informativeness (0.154) user-in-degree (0.133)
4 user-num-posts (0.111) user-num-posts (0.104)
5 content-time-in-day (0.089) user-post-rate (0.075)
6 user-post-rate (0.075) user-out-degree (0.056)
7 content-polarity (0.064) content-referral-count (0.030)
8 user-out-degree (0.040) user-age (0.015)
9 content-referral-count (0.038) content-polarity (0.015)
10 content-length (0.020) content-length (0.010)
11 content-readability (0.018) content-complexity (0.004)
12 user-age (0.015) content-noun-count (0.002)
13 content-uppercase-count (0.012) content-readability (0.001)
14 content-noun-count (0.010) content-verb-count (0.001)
15 content-adj-count (0.005) content-adj-count (0.0)
16 content-complexity (0.0) content-informativeness (0.0)
17 content-verb-count (0.0) content-uppercase-count (0.0)
52
53. 7 content-polarity (0.064) content-referral-count (0.030)
8 user-out-degree (0.040) user-age (0.015)
9 content-referral-count (0.038) content-polarity (0.015)
Positive/negative impact of features
10
11
12
content-length (0.020)
content-readability (0.018)
user-age (0.015)
content-length (0.010)
content-complexity (0.004)
content-noun-count (0.002)
13 content-uppercase-count (0.012) content-readability (0.001)
14 content-noun-count (0.010) content-verb-count (0.001)
• What is the correlation between seed posts and features?
15
16
content-adj-count (0.005)
content-complexity (0.0)
content-adj-count (0.0)
content-informativeness (0.0)
17 content-verb-count (0.0) content-uppercase-count (0.0)
Haiti
Union Address
Fig. 3. Contributions of top-5 features to identifying Non-seeds (N ) and Seeds(S).
Upper plots are for the Haiti dataset and the lower plots are for the Union Address 53
dataset.
54. Predicting discussion activity on Twitter
• Reply rates:
– Haiti 1-74 responses, Union Address 1-75 responses
• Compare rankings
– Ground truth vs predicted
• Experiments
– Using Haiti and Union Address datasets
– Evaluate predicted rank k where k={1,5,10,20,50,100)
– Support Vector Regression with user, content, user+content
features
Dataset Training Test size Test Vol Test Vol SD
size Mean
Haiti 980 210 1.664 3.017
Union Address 5,067 1,161 1.761 2.342 54
55. Predicting discussion activity on Twitter
Haiti dataset Union Address dataset
• Content features are key for top ranks
• Use features more important for higher ranks
55
56. Identifying seed posts in Boards.ie
• Used the same features as before
– User features
• In-degree, out-degree, post count, user age, post rate
– Content features
• Post Length, complexity, readability, referral count, time in day,
informativeness, polarity
• New features designed to capture user affinity
– Forum Entropy
• Concentration of forum activity
• Higher entropy = large forum spread
– Forum Likelihood
• Likelihood of forum post given user history
• Combines post history with incoming data
56
57. Experiment for identifying seed posts
• Used all posts from Boards.ie in 2006
• Built features using a 6-month window prior to seed post date
Posts Seeds Non-Seeds Replies Users
1,942,030 90,765 21,800 1,829,465 29,908
• Evaluated a binary classification task
– Is this post a seed post or not?
– Precision, Recall, F1 and Accuracy
– Tested: user, content, focus features, and their combinations
57
58. h the features (i.e., user TABLE II
om t − 188 to t − 1. In R ESULTS FROMTHE CLASSIFICATION OF SEED POSTS USING
Identifying seeds with different type of
he features compiled for
outcomes and will not
VARYING FEATURE SETS AND CLASSIFICATION MODELS
features
user may increase their
User SVM
P
0.775
R
0.810
F
0.774
ROC
0.581
1
ich would not be a true Naive Bayes 0.691 0.767 0.719 0.540
ime the post was made. Max Ent 0.776 0.806 0.722 0.556
J48 0.778 0.809 0.734 0.582
e number of posts (seeds, Content SVM 0.739 0.804 0.729 0.511
tained within. Naive Bayes 0.730 0.794 0.740 0.616
Max Ent 0.758 0.806 0.730 0.678
TING S EED P OSTS J48 0.795 0.822 0.783 0.617
ls are often hindered by Focus SVM 0.649 0.805 0.719 0.500
Naive Bayes 0.710 0.737 0.722 0.588
We alleviate this problem Max Ent 0.649 0.805 0.719 0.586
and non-seeds through a J48 0.649 0.805 0.719 0.500
posts have been identified User + Content SVM 0.790 0.808 0.727 0.509
Naive Bayes 0.712 0.772 0.732 0.593
of discussion that such Max Ent 0.767 0.807 0.734 0.671
ook for the best classifier J48 0.795 0.821 0.779 0.675
ts and then search for the User + Focus SVM 0.776 0.810 0.776 0.583
Naive Bayes 0.699 0.778 0.724 0.585
guishing seed posts from Max Ent 0.771 0.806 0.722 0.607
atures that are associated J48 0.777 0.810 0.742 0.617
Content + Focus SVM 0.750 0.805 0.729 0.511
Naive Bayes 0.732 0.787 0.746 0.658
Max Ent 0.762 0.807 0.731 0.692
J48 0.798 0.823 0.787 0.662
the previously described All SVM 0.791 0.808 0.727 0.510
ntaining both seeds and Naive Bayes 0.724 0.780 0.740 0.637
Max Ent 0.768 0.808 0.733 0.688
r collection of posts we J48 0.798 0.824 0.792 0.692
tures listed in section III 58
59. Positive/negative impact of features on Boards.ie
TABLE III
R EDUCTION IN F1 LEVELS AS INDIVIDUAL FEATURES ARE
DROPPED FROM THE J 48 CLASSIFIER
• What are the most
Feature Dropped F1
important features for - 0.815
predicting seed posts? Post Count
In-Degree
0.815
0.811*
Out-Degree 0.811*
User Age 0.807***
Post Rate 0.815
Forum Entropy 0.815
• Correlations: Forum Likelihood 0.798***
Post Length 0.810**
– Referral counts (non-seeds) Complexity 0.811**
– Forum likelihood (seeds) Readability 0.802***
Referral Count 0.793***
– Informativeness (non-seeds) Time in Day 0.810**
Informativeness 0.801***
– Readability (seeds) Polarity 0.808***
Signif. codes: p-value < 0.001 *** 0.01 ** 0.05 * 0.1 .
– User age (non-seeds)
hyperlinks (e.g., ads and spams). This contrasts with work in
Twitter which found that tweets containing many links were
59
60. Predicting Discussion Activity in Boards.ie
• What impact do features have on discussion length?
– Assessed Linear Regression model with focus and content
features
– Forum Likelihood (pos)
– Content Length (+/neutral)
– Complexity (pos)
– Readability (+/neutral)
– Referral Count (neg)
– Time in Day (+/neutral)
– Informativeness (-/neutral)
– Polarity (neg)
60
61. Stay tuned
• More communities
– SAP, IBM, StackOverflow, Reddit
– Compare impact of features on their dynamics
• Better behaviour analysis
– Less features, more forums/communities, more graphs!
– Healthy? posts, reciprocation, discussions, sentiment mixture
• Churn analysis
– Correlation of features/behaviour to ‘bounce rate’ (WebSci11 best paper)
• Intervention!
– Opportunities and mechanisms to influence behaviour
61
62. Upcoming events
Social Object Networks
IEEE Social Computing, 2011
October 9-10, Boston, USA
http://ir.ii.uam.es/socialobjects2011/
!
Deadline: August 5, 2011
Intelligent Web Services Meet Social Computing
AAAI Spring Symposium 2012,
March 26-28, Stanford, California
http://vitvar.com/events/aaai-ss12
Deadline: Octover 7, 2011
62
63. Acknowledgement
My social semantics team Live Social Semantics team
Sofia Angeletou Ciro Cattuto Wouter van Den Broeck
Matthew Rowe
Research Associate ISI, Turin ISI, Turin
Research Associate
Alain Barrat Martin Szomszor
CPT Marseille & ISI CeRC, City University, UK
Gianluca Correndo, Uni Southampton
Ivan Cantador, UAM, Madrid
STI International
ESWC09/10 & HT09 chairs and organisers
All LSS participants
63
