Upload
andrea-varga
View
1.118
Download
0
Tags:
Embed Size (px)
DESCRIPTION
The rapid rate of information propagation on social streams has proven to be an up-to-date channel of communication, which can reveal events happening in the world. However, identifying the topicality of short messages (e.g. tweets) distributed on these streams poses new challenges in the development of accurate classification algorithms. In order to alleviate this problem we study for the first time a transfer learning setting aiming to make use of two frequently updated social knowledge sources KSs (DBpedia and Freebase) for detecting topics in tweets. In this paper we investigate the similarity (and dissimilarity) between these KSs and Twitter at the lexical and conceptual (entity) level. We also evaluate the contribution of these types of features and propose various statistical measures for determining the topics which are highly similar or different in KSs and tweets. Our findings can be of potential use to machine learning or domain adaptation algorithms aiming to use named entities for topic classification of tweets. These results can also be valuable in the identification of representative sets of annotated articles from the KSs, which can help in building accurate topic classifiers of tweets.
Citation preview
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Exploring the similarity between Social Knowledge Sources andTwitter for Cross-Domain Topic Classification of Tweets
Andrea Varga, Amparo E. Cano and Fabio Ciravegna
1 Organisations Information and Knowledge (OAK) Research GroupUniversity of Sheffield
2 Knowledge Management Institute (KMI)Open University
KECSM 2012/ISWC 2012
Nov 12, 2012
1/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Outline
1 Motivation
2 State-of-the-art
3 Methodology
4 Results
5 Conclusions and Future Work
2/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Why classifying Tweets into topics?
Topic classification (TC) of tweets can be important for multipleapplication:
Information Retrieval
Recommendation
Emergency responses, etc.
Topic name Example tweetsDisaster&Accident(DisAcc) happening accident people dying could
phone ambulance wakakkaka xdEntertainment&Culture(EntCult) google adwords commercial greeeat en-
joyed watching greeeeeat dayPolitics(Pol) quoting military source sk media reports
deployed rocket launchers decoys realSports(Sports) ravens good position games left browns
bengals playoffs
3/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
What are the challenges in Topic Classification (TC) of Tweets?
Special characteristics of tweets
the restricted size of a post (limited to 140 characters)
the frequent use of misspellings and jargons
the frequent use of abbreviations
the use of non-standard English: reflected in vocabulary and writing style
Topic name Example tweetsDisaster&Accident(DisAcc) happening accident people dying could
phone ambulance wakakkaka xdEntertainment&Culture(EntCult) google commercial greeeat enjoyed
watching dayPolitics(Pol) quoting military source media reports de-
ployed rocket launchers decoys realSports(Sports) ravens good position games left browns
bengals playoffs
4/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
What are the challenges in Topic Classification (TC) of Tweets?
Special characteristics of tweets
the restricted size of a post (limited to 140 characters)
the frequent use of misspellings and jargons
the frequent use of abbreviations
the use of non-standard English: reflected in vocabulary and writing style
Topic name Example tweetsDisaster&Accident(DisAcc) happening accident people dying could
phone ambulance wakakkaka xdEntertainment&Culture(EntCult) google adwords commercial greeeat en-
joyed watching greeeeeat dayPolitics(Pol) quoting military source media reports de-
ployed rocket launchers decoys realSports(Sports) ravens good position games left browns
bengals playoffs
4/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
What are the challenges in Topic Classification (TC) of Tweets?
Special characteristics of tweets
the restricted size of a post (limited to 140 characters)
the frequent use of misspellings and jargons
the frequent use of abbreviations
the use of non-standard English: reflected in vocabulary and writing style
Topic name Example tweetsDisaster&Accident(DisAcc) happening accident people dying could
phone ambulance wakakkaka xdEntertainment&Culture(EntCult) google adwords commercial greeeat en-
joyed watching greeeeeat dayPolitics(Pol) quoting military source sk media reports
deployed rocket launchers decoys realSports(Sports) ravens good position games left browns
bengals playoffs
4/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
What are the challenges in Topic Classification (TC) of Tweets?
Special characteristics of tweets
the restricted size of a post (limited to 140 characters)
the frequent use of misspellings and jargons
the frequent use of abbreviations
the use of non-standard English: reflected in vocabulary and writing style
Topic name Example tweetsDisaster&Accident(DisAcc) happening accident people dying could
phone ambulance wakakkaka xdEntertainment&Culture(EntCult) google adwords commercial greeeat en-
joyed watching greeeeeat dayPolitics(Pol) quoting military source sk media reports
deployed rocket launchers decoys realSports(Sports) ravens good position games left browns
bengals playoffs
4/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
What are the challenges in Topic Classification (TC) of Tweets?
Special characteristics of tweets
the restricted size of a post (limited to 140 characters)
the frequent use of misspellings and jargons
the frequent use of abbreviations
the use of non-standard English: reflected in vocabulary and writing style
Topic name Example tweetsDisaster&Accident(DisAcc) happening accident people dying could
phone ambulance wakakkaka xdEntertainment&Culture(EntCult) google adwords commercial greeeat en-
joyed watching greeeeeat dayPolitics(Pol) quoting military source sk media reports
deployed rocket launchers decoys realSports(Sports) ravens good position games left browns
bengals playoffs
=> These characteristics poses additional challenges for traditionalsupervised machine learning approaches for buildingaccurate TC of tweets
4/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Why are Social Knowledge Sources (KS) relevant to Twitter?
Data bottleneck problem: investigate an alternative approach inspiredby domain adaptation/transfer learning for exploiting the information fromSocial Knowledge Sources (DBpedia and Freebase) for TC of Tweets
Commonalities between KSs and Twitter
they are constantly edited by web users
they are social and built on a collaborative manner
they cover a large number of topics
5/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Why are Social Knowledge Sources (KS) relevant to Twitter?
Data bottleneck problem: investigate an alternative approach inspiredby domain adaptation/transfer learning for exploiting the information fromSocial Knowledge Sources (DBpedia and Freebase) for TC of Tweets
Commonalities between KSs and Twitter
they are constantly edited by web users
they are social and built on a collaborative manner
they cover a large number of topics
5/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Why are Social Knowledge Sources (KS) relevant to Twitter?
Data bottleneck problem: investigate an alternative approach inspiredby domain adaptation/transfer learning for exploiting the information fromSocial Knowledge Sources (DBpedia and Freebase) for TC of Tweets
Commonalities between KSs and Twitter
they are constantly edited by web users
they are social and built on a collaborative manner
they cover a large number of topics
5/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Why are Social Knowledge Sources (KS) relevant to Twitter?
Data bottleneck problem: investigate an alternative approach inspiredby domain adaptation/transfer learning for exploiting the information fromSocial Knowledge Sources (DBpedia and Freebase) for TC of Tweets
Commonalities between KSs and Twitter
they are constantly edited by web users
they are social and built on a collaborative manner
they cover a large number of topics
5/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Why are Social Knowledge Sources (KS) relevant to Twitter?
Data bottleneck problem: investigate an alternative approach inspiredby domain adaptation/transfer learning for exploiting the information fromSocial Knowledge Sources (DBpedia and Freebase) for TC of Tweets
Commonalities between KSs and Twitter
they are constantly edited by web users
they are social and built on a collaborative manner
they cover a large number of topics
More importantly: KSs contain a large number of annotated data on alarge number of topics
5/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Research questions
1 Are KSs relevant for topic classification of Tweets?
2 Which features make the KSs look more similar to Twitter?
3 How similar or dissimilar are KSs to Twitter? Which similarity measuredoes better quantify the lexical changes between KSs and Twitter?
6/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
State-of-the-art approaches for TC of Tweets
Using DBpedia for Topic Classification of Tweets:Wikify (Mihalcea, R. and Csomai, A., 2007)Enriching unstructured text with Wikipedia links (D. Milne and I. H. Witten,2008)Tagme (P. Ferragina and U. Scaiella., 2010)Topical Social Sensor (P. K. P. N. Mendes et al., 2010)Vector space model (Oscar Munoz-Garcia et al. 2011)
Using Freebase for Topic Classification of Tweets:Clustering based approach (S.P.Kasiviswanathan et al., 2011)
Our main contribution:Understanding the similarity between KSs and Twitter
Exploring multiple KSs (DBpedia + Freebase)
Investigating various statistical metrics for quantifying thesimilarity between KSs and Twitter
7/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Methodology followed
8/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Methodology followed
1 Collecting Data from KSs
Retrieve articles
Concept enrichment
Build Cross-domain Classifier Annotate Tweets
Sc. FB Sc. DB-FBSc. DB
Retrieve tweets
Concept enrichment
8/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Methodology followed
1 Collecting Data from KSs
2 Building Cross-Domain (CD) Topic Classifier of Tweets
Retrieve articles
Concept enrichment
Build Cross-domain Classifier Annotate Tweets
Sc. FB Sc. DB-FBSc. DB
Retrieve tweets
Concept enrichment
8/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Methodology followed
1 Collecting Data from KSs
2 Building Cross-Domain (CD) Topic Classifier of Tweets
3 Measuring Distributional Changes Between KSs and Twitter
Retrieve articles
Concept enrichment
Build Cross-domain Classifier Annotate Tweets
Sc. FB Sc. DB-FBSc. DB
Retrieve tweets
Concept enrichment
8/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Step 1: Collecting Data from KSs
Twitter corpus collected in Abel et al. (2011), tweets posted between October 2010 andJanuary 2011, annotated with 17 topics
Random selection of 1,000 articles/tweets from DBpedia/Freebase/Twitter for each topic =>9,465 articles from DBpedia; 16,915 articles from Freebase; and 12,412 tweets
Preprocessing: removal of hastags, mentions and URLs from tweets; taking top-1000
features for each topic
88.6%
8.6%
1.8%0.9%
Dbpedia multilabel frequency
1 8 2 3+4+5+6+7+9
99.9% 0.1%
Freebase multilabel frequency
1 2
71%
22.3%
5.6%
1%0.1%
Twitter multilabel frequency
1 2 3 4 6+5
88.6%
8.6%
1.8%0.9%
Dbpedia multilabel frequency
1 8 2 3+4+5+6+7+9
99.9% 0.1%
Freebase multilabel frequency
1 2
71%
22.3%
5.6%
1%0.1%
Twitter multilabel frequency
1 2 3 4 6+5
9/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Step 1: Collecting Data from KSs
Disaster_AccidentBusiness_Finance Education Entertainment
Health
Environment
Human Interest Labor Law_Crime
Politics
Religion Social Issues Sports
Technology_IT
Weather War_Conflict
Retrieval of articles for a given topic (e.g. Politics):from DBpedia: executing SPARQL queries for retrieving category namescontaining the topic name:
Category:Politics_of_the_United_StatesCategory:National_Democratic_Party_Egypt_politiciansetc.
from Freebase: accessing Text Service API for articles belonging to thetopic:
for underspecified topics/domains: consider articles containing the topic in theirtitles
10/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Step 2: Building Cross-Domain (CD) Topic Classifier of Tweets
Considering two different feature sets:
11/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Step 2: Building Cross-Domain (CD) Topic Classifier of Tweets
Considering two different feature sets:BOW: tf.idf value of the words present the examples (articles or tweets)
11/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Step 2: Building Cross-Domain (CD) Topic Classifier of Tweets
Considering two different feature sets:BOW: tf.idf value of the words present the examples (articles or tweets)
BOE: tf.idf value of the words and entity+concept pairs present the examples(articles or tweets)
Retrieve articles
Concept enrichment
Build Cross-domain Classifier Annotate Tweets
Sc. FB Sc. DB-FBSc. DB
Retrieve tweets
Concept enrichment
11/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Step 3: Measuring Distributional Changes Between KSs and Twitter
Building a vector−→ds for each the source dataset (Sc.DB, Sc.Fb,
Sc.Db-FB) and a vector−→dt for the target dataset (Twitter) consisting of
the TF-IDF weight for either the BoW or BoE feature sets
statistical measures applied:
χ2 test: χ2 =∑ (O−E)2
E , where O is the observed value for a feature, whileE is the expected value calculated on the basis of the joint corpus
Kullback-Leibler symmetric distance:
KL(−→ds ||−→dt ) =
∑f∈FS∪FT
(−→ds(f )−
−→dt (f )) log
−→ds (f )−→dt (f )
cosine similarity: cosine(−→ds ,−→dt ) =
∑FS∪FTk=1 (
−→ds (fSk
)×−→ds (fTk
))∑FS∪FTk=1 (
−→ds (fSk
))2×∑FS∪FT
k=1 (−→dt (fTk
))2
12/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Experimental setting
1-vs-all approach, building individual CD classifier for each topic, SVMclassifiers, performed 5 cross-fold validation
Sc-Db, Sc-Fb, Sc-Db-Fb classifiers trained on full KS data, evaluated on20% Twitter data 2,482 tweets)
TGT classifier: trained on 80% Twitter data, evaluated on 20% Twitterdata (2,482 tweets)
13/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q1 : Which KS reflects better the lexical variation in Twitter?
14/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q1 : Which KS reflects better the lexical variation in Twitter?
Sc.D
B(Bo
E)
Sc.D
B(Bo
W)
Sc.F
B(Bo
E)
SC.F
B(Bo
W)
Sc.D
B−FB
(BoE
)
SC.D
B−FB
(BoW
)
TGT(
BoW
)
TGT(
BoE)
Edu
Sports
War
Labor
Weather
HumInt
Env
TechIT
DisAcc
SocIssue
HospRecr
Law
Pol
Health
Religion
EntCult
BusFi
37.40 37.80 42.50 47.20 42.50 45.70 71.90 71.30
10.30 11.70 26.50 26.20 26.20 26.00 60.10 59.20
9.30 10.90 23.90 23.60 24.60 25.70 67.60 72.70
1.40 1.30 31.90 31.90 30.10 29.90 79.90 79.40
3.10 7.00 39.80 39.90 36.70 36.00 81.20 81.50
1.10 1.40 2.00 1.60 1.50 2.20 33.60 34.20
15.20 14.20 2.20 2.20 8.90 8.40 46.60 48.30
1.60 2.40 18.40 18.60 12.40 9.90 57.40 58.00
8.30 9.70 14.50 14.50 21.00 21.10 57.50 59.20
1.30 2.00 8.80 9.00 9.70 9.10 44.20 44.00
1.40 2.30 17.20 19.50 11.30 13.90 41.60 42.60
0.90 2.70 16.80 17.80 14.80 13.30 46.80 46.40
22.20 21.00 27.80 26.80 24.20 26.80 45.10 44.90
28.60 30.30 25.60 24.70 26.90 25.40 51.70 51.80
23.40 25.10 14.40 14.70 21.00 20.40 58.40 58.50
15.30 15.10 15.50 16.20 19.50 20.20 42.20 43.10
21.40 21.20 11.50 15.30 18.60 19.10 47.50 46.50
14/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q1 : Which KS reflects better the lexical variation in Twitter?
Sc.D
B(Bo
E)
Sc.D
B(Bo
W)
Sc.F
B(Bo
E)
SC.F
B(Bo
W)
Sc.D
B−FB
(BoE
)
SC.D
B−FB
(BoW
)
TGT(
BoW
)
TGT(
BoE)
Edu
Sports
War
Labor
Weather
HumInt
Env
TechIT
DisAcc
SocIssue
HospRecr
Law
Pol
Health
Religion
EntCult
BusFi
37.40 37.80 42.50 47.20 42.50 45.70 71.90 71.30
10.30 11.70 26.50 26.20 26.20 26.00 60.10 59.20
9.30 10.90 23.90 23.60 24.60 25.70 67.60 72.70
1.40 1.30 31.90 31.90 30.10 29.90 79.90 79.40
3.10 7.00 39.80 39.90 36.70 36.00 81.20 81.50
1.10 1.40 2.00 1.60 1.50 2.20 33.60 34.20
15.20 14.20 2.20 2.20 8.90 8.40 46.60 48.30
1.60 2.40 18.40 18.60 12.40 9.90 57.40 58.00
8.30 9.70 14.50 14.50 21.00 21.10 57.50 59.20
1.30 2.00 8.80 9.00 9.70 9.10 44.20 44.00
1.40 2.30 17.20 19.50 11.30 13.90 41.60 42.60
0.90 2.70 16.80 17.80 14.80 13.30 46.80 46.40
22.20 21.00 27.80 26.80 24.20 26.80 45.10 44.90
28.60 30.30 25.60 24.70 26.90 25.40 51.70 51.80
23.40 25.10 14.40 14.70 21.00 20.40 58.40 58.50
15.30 15.10 15.50 16.20 19.50 20.20 42.20 43.10
21.40 21.20 11.50 15.30 18.60 19.10 47.50 46.50
14/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q1 : Which KS reflects better the lexical variation in Twitter?
Sc.D
B(Bo
E)
Sc.D
B(Bo
W)
Sc.F
B(Bo
E)
SC.F
B(Bo
W)
Sc.D
B−FB
(BoE
)
SC.D
B−FB
(BoW
)
TGT(
BoW
)
TGT(
BoE)
Edu
Sports
War
Labor
Weather
HumInt
Env
TechIT
DisAcc
SocIssue
HospRecr
Law
Pol
Health
Religion
EntCult
BusFi
37.40 37.80 42.50 47.20 42.50 45.70 71.90 71.30
10.30 11.70 26.50 26.20 26.20 26.00 60.10 59.20
9.30 10.90 23.90 23.60 24.60 25.70 67.60 72.70
1.40 1.30 31.90 31.90 30.10 29.90 79.90 79.40
3.10 7.00 39.80 39.90 36.70 36.00 81.20 81.50
1.10 1.40 2.00 1.60 1.50 2.20 33.60 34.20
15.20 14.20 2.20 2.20 8.90 8.40 46.60 48.30
1.60 2.40 18.40 18.60 12.40 9.90 57.40 58.00
8.30 9.70 14.50 14.50 21.00 21.10 57.50 59.20
1.30 2.00 8.80 9.00 9.70 9.10 44.20 44.00
1.40 2.30 17.20 19.50 11.30 13.90 41.60 42.60
0.90 2.70 16.80 17.80 14.80 13.30 46.80 46.40
22.20 21.00 27.80 26.80 24.20 26.80 45.10 44.90
28.60 30.30 25.60 24.70 26.90 25.40 51.70 51.80
23.40 25.10 14.40 14.70 21.00 20.40 58.40 58.50
15.30 15.10 15.50 16.20 19.50 20.20 42.20 43.10
21.40 21.20 11.50 15.30 18.60 19.10 47.50 46.50
14/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q1 : Which KS reflects better the lexical variation in Twitter?
Sc.D
B(Bo
E)
Sc.D
B(Bo
W)
Sc.F
B(Bo
E)
SC.F
B(Bo
W)
Sc.D
B−FB
(BoE
)
SC.D
B−FB
(BoW
)
TGT(
BoW
)
TGT(
BoE)
Edu
Sports
War
Labor
Weather
HumInt
Env
TechIT
DisAcc
SocIssue
HospRecr
Law
Pol
Health
Religion
EntCult
BusFi
37.40 37.80 42.50 47.20 42.50 45.70 71.90 71.30
10.30 11.70 26.50 26.20 26.20 26.00 60.10 59.20
9.30 10.90 23.90 23.60 24.60 25.70 67.60 72.70
1.40 1.30 31.90 31.90 30.10 29.90 79.90 79.40
3.10 7.00 39.80 39.90 36.70 36.00 81.20 81.50
1.10 1.40 2.00 1.60 1.50 2.20 33.60 34.20
15.20 14.20 2.20 2.20 8.90 8.40 46.60 48.30
1.60 2.40 18.40 18.60 12.40 9.90 57.40 58.00
8.30 9.70 14.50 14.50 21.00 21.10 57.50 59.20
1.30 2.00 8.80 9.00 9.70 9.10 44.20 44.00
1.40 2.30 17.20 19.50 11.30 13.90 41.60 42.60
0.90 2.70 16.80 17.80 14.80 13.30 46.80 46.40
22.20 21.00 27.80 26.80 24.20 26.80 45.10 44.90
28.60 30.30 25.60 24.70 26.90 25.40 51.70 51.80
23.40 25.10 14.40 14.70 21.00 20.40 58.40 58.50
15.30 15.10 15.50 16.20 19.50 20.20 42.20 43.10
21.40 21.20 11.50 15.30 18.60 19.10 47.50 46.50
14/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q1 : Which KS reflects better the lexical variation in Twitter?
Sc.Db-FB showed best performance, followed by Sc.Fb and Sc.Db
Sc.D
B(Bo
E)
Sc.D
B(Bo
W)
Sc.F
B(Bo
E)
SC.F
B(Bo
W)
Sc.D
B−FB
(BoE
)
SC.D
B−FB
(BoW
)
TGT(
BoW
)
TGT(
BoE)
Edu
Sports
War
Labor
Weather
HumInt
Env
TechIT
DisAcc
SocIssue
HospRecr
Law
Pol
Health
Religion
EntCult
BusFi
37.40 37.80 42.50 47.20 42.50 45.70 71.90 71.30
10.30 11.70 26.50 26.20 26.20 26.00 60.10 59.20
9.30 10.90 23.90 23.60 24.60 25.70 67.60 72.70
1.40 1.30 31.90 31.90 30.10 29.90 79.90 79.40
3.10 7.00 39.80 39.90 36.70 36.00 81.20 81.50
1.10 1.40 2.00 1.60 1.50 2.20 33.60 34.20
15.20 14.20 2.20 2.20 8.90 8.40 46.60 48.30
1.60 2.40 18.40 18.60 12.40 9.90 57.40 58.00
8.30 9.70 14.50 14.50 21.00 21.10 57.50 59.20
1.30 2.00 8.80 9.00 9.70 9.10 44.20 44.00
1.40 2.30 17.20 19.50 11.30 13.90 41.60 42.60
0.90 2.70 16.80 17.80 14.80 13.30 46.80 46.40
22.20 21.00 27.80 26.80 24.20 26.80 45.10 44.90
28.60 30.30 25.60 24.70 26.90 25.40 51.70 51.80
23.40 25.10 14.40 14.70 21.00 20.40 58.40 58.50
15.30 15.10 15.50 16.20 19.50 20.20 42.20 43.10
21.40 21.20 11.50 15.30 18.60 19.10 47.50 46.50
15/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q1 : Which KS reflects better the lexical variation in Twitter?
Sc.Db-FB showed best performance, followed by Sc.Fb and Sc.Db
Sc.D
B(Bo
E)
Sc.D
B(Bo
W)
Sc.F
B(Bo
E)
SC.F
B(Bo
W)
Sc.D
B−FB
(BoE
)
SC.D
B−FB
(BoW
)
TGT(
BoW
)
TGT(
BoE)
Edu
Sports
War
Labor
Weather
HumInt
Env
TechIT
DisAcc
SocIssue
HospRecr
Law
Pol
Health
Religion
EntCult
BusFi
37.40 37.80 42.50 47.20 42.50 45.70 71.90 71.30
10.30 11.70 26.50 26.20 26.20 26.00 60.10 59.20
9.30 10.90 23.90 23.60 24.60 25.70 67.60 72.70
1.40 1.30 31.90 31.90 30.10 29.90 79.90 79.40
3.10 7.00 39.80 39.90 36.70 36.00 81.20 81.50
1.10 1.40 2.00 1.60 1.50 2.20 33.60 34.20
15.20 14.20 2.20 2.20 8.90 8.40 46.60 48.30
1.60 2.40 18.40 18.60 12.40 9.90 57.40 58.00
8.30 9.70 14.50 14.50 21.00 21.10 57.50 59.20
1.30 2.00 8.80 9.00 9.70 9.10 44.20 44.00
1.40 2.30 17.20 19.50 11.30 13.90 41.60 42.60
0.90 2.70 16.80 17.80 14.80 13.30 46.80 46.40
22.20 21.00 27.80 26.80 24.20 26.80 45.10 44.90
28.60 30.30 25.60 24.70 26.90 25.40 51.70 51.80
23.40 25.10 14.40 14.70 21.00 20.40 58.40 58.50
15.30 15.10 15.50 16.20 19.50 20.20 42.20 43.10
21.40 21.20 11.50 15.30 18.60 19.10 47.50 46.50
15/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q1 : Which KS reflects better the lexical variation in Twitter?
Sc.Db-FB showed best performance, followed by Sc.Fb and Sc.Db
Sc.D
B(Bo
E)
Sc.D
B(Bo
W)
Sc.F
B(Bo
E)
SC.F
B(Bo
W)
Sc.D
B−FB
(BoE
)
SC.D
B−FB
(BoW
)
TGT(
BoW
)
TGT(
BoE)
Edu
Sports
War
Labor
Weather
HumInt
Env
TechIT
DisAcc
SocIssue
HospRecr
Law
Pol
Health
Religion
EntCult
BusFi
37.40 37.80 42.50 47.20 42.50 45.70 71.90 71.30
10.30 11.70 26.50 26.20 26.20 26.00 60.10 59.20
9.30 10.90 23.90 23.60 24.60 25.70 67.60 72.70
1.40 1.30 31.90 31.90 30.10 29.90 79.90 79.40
3.10 7.00 39.80 39.90 36.70 36.00 81.20 81.50
1.10 1.40 2.00 1.60 1.50 2.20 33.60 34.20
15.20 14.20 2.20 2.20 8.90 8.40 46.60 48.30
1.60 2.40 18.40 18.60 12.40 9.90 57.40 58.00
8.30 9.70 14.50 14.50 21.00 21.10 57.50 59.20
1.30 2.00 8.80 9.00 9.70 9.10 44.20 44.00
1.40 2.30 17.20 19.50 11.30 13.90 41.60 42.60
0.90 2.70 16.80 17.80 14.80 13.30 46.80 46.40
22.20 21.00 27.80 26.80 24.20 26.80 45.10 44.90
28.60 30.30 25.60 24.70 26.90 25.40 51.70 51.80
23.40 25.10 14.40 14.70 21.00 20.40 58.40 58.50
15.30 15.10 15.50 16.20 19.50 20.20 42.20 43.10
21.40 21.20 11.50 15.30 18.60 19.10 47.50 46.50
15/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Classification performance in F1 measure
Q2 : What feature makes the KSs look more similar to Twitter?
BoW features were found better than BoE for CD classifiersBoE features were found better than BoW for TGT
Sc.D
B(Bo
E)
Sc.D
B(Bo
W)
Sc.F
B(Bo
E)
SC.F
B(Bo
W)
Sc.D
B−FB
(BoE
)
SC.D
B−FB
(BoW
)
TGT(
BoW
)
TGT(
BoE)
Edu
Sports
War
Labor
Weather
HumInt
Env
TechIT
DisAcc
SocIssue
HospRecr
Law
Pol
Health
Religion
EntCult
BusFi
37.40 37.80 42.50 47.20 42.50 45.70 71.90 71.30
10.30 11.70 26.50 26.20 26.20 26.00 60.10 59.20
9.30 10.90 23.90 23.60 24.60 25.70 67.60 72.70
1.40 1.30 31.90 31.90 30.10 29.90 79.90 79.40
3.10 7.00 39.80 39.90 36.70 36.00 81.20 81.50
1.10 1.40 2.00 1.60 1.50 2.20 33.60 34.20
15.20 14.20 2.20 2.20 8.90 8.40 46.60 48.30
1.60 2.40 18.40 18.60 12.40 9.90 57.40 58.00
8.30 9.70 14.50 14.50 21.00 21.10 57.50 59.20
1.30 2.00 8.80 9.00 9.70 9.10 44.20 44.00
1.40 2.30 17.20 19.50 11.30 13.90 41.60 42.60
0.90 2.70 16.80 17.80 14.80 13.30 46.80 46.40
22.20 21.00 27.80 26.80 24.20 26.80 45.10 44.90
28.60 30.30 25.60 24.70 26.90 25.40 51.70 51.80
23.40 25.10 14.40 14.70 21.00 20.40 58.40 58.50
15.30 15.10 15.50 16.20 19.50 20.20 42.20 43.10
21.40 21.20 11.50 15.30 18.60 19.10 47.50 46.50
16/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Examining the number of annotation needed for Twitterclassifier to outperform Sc. Db-FB
Investigated the impact of employing Sc. Db-FB classifier over theTwitter classifier in terms of number of annotations
The performance of the Twitter classifier against the three CD classifiersover the full learning curve
17/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Examining the number of annotation needed for Twitterclassifier to outperform Sc. Db-FB
Investigated the impact of employing Sc. Db-FB classifier over theTwitter classifier in terms of number of annotations
The performance of the Twitter classifier against the three CD classifiersover the full learning curve
=> In the absence of any annotated tweets, applying these CDclassifiers are beneficial
17/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Examining the number of annotation needed for Twitterclassifier to outperform the CD classifiers
Q3 : How similar or dissimilar are KSs to Twitter posts; and whichsimilarity measure does better reflect the lexical changes between KSsand Twitter posts?
Compared χ2, KL-divergence, cosine for each topic
χ2 obtained the best correlation with the performance of CD classifiers,achived scores >70% for 32 cases
cosine obtained correlation scores >70% for 25 cases
KL obtained correlation scores >70% for 24 cases
18/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Findings -Examining the number of annotation needed for Twitterclassifier to outperform the CD classifiers
Q3 : How similar or dissimilar are KSs to Twitter posts; and whichsimilarity measure does better reflect the lexical changes between KSsand Twitter posts?
Compared χ2, KL-divergence, cosine for each topic
χ2 obtained the best correlation with the performance of CD classifiers,achived scores >70% for 32 cases
cosine obtained correlation scores >70% for 25 cases
KL obtained correlation scores >70% for 24 cases
=> χ2 test is the best measure for quantifying the distributionaldifferences between KSs and Twitter.
18/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Conclusions and Future Work
We presented a first study towards understanding the usefulness of KSs in TC oftweets at various granularities: lexical features (BoW) and entity features (BoE)
Our main findings are:
19/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Conclusions and Future Work
We presented a first study towards understanding the usefulness of KSs in TC oftweets at various granularities: lexical features (BoW) and entity features (BoE)
Our main findings are:In the absence of any annotated tweets, applying these CD classifiers are beneficial
19/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Conclusions and Future Work
We presented a first study towards understanding the usefulness of KSs in TC oftweets at various granularities: lexical features (BoW) and entity features (BoE)
Our main findings are:In the absence of any annotated tweets, applying these CD classifiers are beneficial
Out of the two KSs, Freebase topics seem to be much closer to the Twitter topics thanthe DBpedia topics.
19/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Conclusions and Future Work
We presented a first study towards understanding the usefulness of KSs in TC oftweets at various granularities: lexical features (BoW) and entity features (BoE)
Our main findings are:In the absence of any annotated tweets, applying these CD classifiers are beneficial
Out of the two KSs, Freebase topics seem to be much closer to the Twitter topics thanthe DBpedia topics.
The two KSs contain complementary information
19/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Conclusions and Future Work
We presented a first study towards understanding the usefulness of KSs in TC oftweets at various granularities: lexical features (BoW) and entity features (BoE)
Our main findings are:In the absence of any annotated tweets, applying these CD classifiers are beneficial
Out of the two KSs, Freebase topics seem to be much closer to the Twitter topics thanthe DBpedia topics.
The two KSs contain complementary information
For the CD classifiers, on average BOW features were more useful than BoE features
19/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Conclusions and Future Work
We presented a first study towards understanding the usefulness of KSs in TC oftweets at various granularities: lexical features (BoW) and entity features (BoE)
Our main findings are:In the absence of any annotated tweets, applying these CD classifiers are beneficial
Out of the two KSs, Freebase topics seem to be much closer to the Twitter topics thanthe DBpedia topics.
The two KSs contain complementary information
For the CD classifiers, on average BOW features were more useful than BoE features
For the Twitter classifiers, on average BOE features were more useful than BoWfeatures
19/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Conclusions and Future Work
We presented a first study towards understanding the usefulness of KSs in TC oftweets at various granularities: lexical features (BoW) and entity features (BoE)
Our main findings are:In the absence of any annotated tweets, applying these CD classifiers are beneficial
Out of the two KSs, Freebase topics seem to be much closer to the Twitter topics thanthe DBpedia topics.
The two KSs contain complementary information
For the CD classifiers, on average BOW features were more useful than BoE features
For the Twitter classifiers, on average BOE features were more useful than BoWfeatures
We found χ2 test as being the best measure for quantifying the distributionaldifferences between KSs and Twitter.
19/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Conclusions and Future Work
We presented a first study towards understanding the usefulness of KSs in TC oftweets at various granularities: lexical features (BoW) and entity features (BoE)
Our main findings are:In the absence of any annotated tweets, applying these CD classifiers are beneficial
Out of the two KSs, Freebase topics seem to be much closer to the Twitter topics thanthe DBpedia topics.
The two KSs contain complementary information
For the CD classifiers, on average BOW features were more useful than BoE features
For the Twitter classifiers, on average BOE features were more useful than BoWfeatures
We found χ2 test as being the best measure for quantifying the distributionaldifferences between KSs and Twitter.
Our future work will focus on building more accurate TC classifiers andinvestigating better measures
19/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Corpus Analysis - Size of vocabulary
!"
#!!!"
$!!!"
%!!!"
&!!!"
'!!!!"
'#!!!"
'$!!!"
'%!!!"
()*+,"
-,*.//"
01)"
0234)53"
026"
78953:"
7;*<=8/"
7)>?23"
@9A;B"
@9C"
D;5"
=85,E,;2"
F;/?**)8"
F<;B3*"
G8/:?G"
H893:8B"
H9B"
IJ%'" I%&!"IIJ$" IIJI"
I&%!"I$K!" I$L$" I#%J" I'!!"
IJ#I" I$!I" I#&#" I#JK"IJL'" II%I" IIK!" I'$#"
&K%&"LI$#"
K$'#"
'!#'#"
LJJ$"LI!K" LI$#" L$##" LI$#" LI$#"
LK'&"
''L$I"
LI$#"
&'&I"
L&KJ"
&&J&"LI$#"
I&!I"
#'&L"
IK!K"
##J$"'&#J"
$%&$"
JK%#"
%JK'"
'J%&"
$JJ'"
%$'#"J&&&"
'&LK"
%&$#" %K#&"
#$#K"#KJ!"
'!K#'"
'!!L#"
L##'"
'''!!"
'!!I'"
''L%L"'#$#J"
'I!$L"
L&#I"
''J#I"
'IIIJ"
'J!I'"
'!!#L"
'#J!K"
'I&'L"
LLII"'!$IJ"
GMG"
F/N-("
F/N+("
F/N-(O+("
20/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Understanding the results - Number of unique entities
Examining the number of entities in the source (Sc. DB, Sc. FB, Sc. DB-FB) andtarget (TGT) datasets after pre-processing.
the TGT dataset consists of 1.73± 0.35 entities/tweet
the Sc.DB dataset consists of 22.24± 1.44 entities entities/article
the Sc.FB dataset consists of 8.14± 5.78 entities entities/article
!"
#!!!"
$!!!!"
$#!!!"
%!!!!"
%#!!!"
&!!!!"
&#!!!"
'!!!!"
'#!!!"
()*+,"
-,*.//"
01)"
0234)53"
026"
78953:"
7;*<=8/"
7)>?23"
@9A;B"
@9C"
D;5"
=85,E,;2"
F;/?**)8"
F<;B3*"
G8/:?G"
H893:8B"
H9B"
$II%" $JKK" $J!#" $J!#" $JK&" $'K$" $%J&" $#I$" $'#&" $ILI" %&$%"$$%!" $#$%" %$$L" $KJ&" %%JL"
$%!L"
%$$#L"%%&K&" %%!IL"
%&'#!"%%%J'"
%!LI%"
%%&K&" %%#K%" %%&K&" %%&K&"%!JII"
$KIK'"
%%&K&"
%'KK$"%&L'$"
%%%&I" %%&K&"
L'!$"
%&$L"
KL%%"
$I$&" %&J#"
#K#J"
IJ#$"
$'L&L"
%'%'"
KIK%"
$I%&!"
$%!#&"
%''$"
$J'K&"
$'$#&"
%J%I"
KIKK"
%K!K#"
%'#%'"
%I$J&"
%#%&L"%'&'#"
%#I!J"
&$LJ%"
&LL'$"
%'#KK"
%IIJK"
&I%JK"
%I&$!"
%'#IL"
'%&KI"
&K&!!"
%'J&&"
%III%"
GMG"
F/N-("
F/N+("
F/N-(O+("
21/22
Motivation Research question State-of-the-art Methodology Results Conclusions and Future Work
Corpus Analysis - Distribution of the top 15 entities
Labo
r (Sc
.FB)
War
(TG
T)W
ar (S
c.FB
)So
cIss
ue (S
c.FB
)W
eath
er (S
c.FB
)Te
chIT
(Sc.
DB)
Hum
Int (
Sc.D
B)W
eath
er (S
c.DB
)So
cIss
ue (S
c.DB
)La
w (S
c.DB
)La
bor (
Sc.D
B)Di
sAcc
(Sc.
DB)
Hosp
Recr
(Sc.
DB)
EntC
ult (
Sc.F
B)Te
chIT
(Sc.
FB)
Hum
Int (
TGT)
EntC
ult (
TGT)
EntC
ult (
Sc.D
B)Te
chIT
(TG
T)Bu
sFi (
Sc.F
B)He
alth
(Sc.
FB)
Heal
th (S
c.DB
)Bu
sFi (
Sc.D
B)En
v (S
c.DB
)La
w (S
c.FB
)En
v (S
c.FB
)Di
sAcc
(Sc.
FB)
Hosp
Recr
(Sc.
FB)
Wea
ther
(TG
T)Di
sAcc
(TG
T)En
v (T
GT)
Edu
(Sc.
DB)
Edu
(Sc.
FB)
Spor
ts (S
c.DB
)W
ar (S
c.DB
)Hu
mIn
t (Sc
.FB)
Pol (
Sc.F
B)Sp
orts
(Sc.
FB)
Relig
ion
(Sc.
FB)
Spor
ts (T
GT)
Relig
ion
(TG
T)Po
l (TG
T)Po
l (Sc
.DB)
Relig
ion
(Sc.
DB)
Heal
th (T
GT)
Hosp
Recr
(TG
T)Bu
sFi (
TGT)
Labo
r (TG
T)So
cIss
ue (T
GT)
Edu
(TG
T)La
w (T
GT)
NaturalFeature
Country
Person
Position
Organization
Technology
MedicalCondition
SportsEvent
Region
Continent
Company
IndustryTerm
City
Facility
ProvinceOrState
22/22