Unsupervised graph based patterns extraction for emotion classification

Unsupervised Graph-Based Patterns Extraction for Emotion

Classification

ASONAM 2015 Conference

o

o

o

o

ASONAM 2015 Conference

ASONAM 2015 Conference

●●

●o

●o …

ASONAM 2015 Conference

ASONAM 2015 Conference

● Kim et al. (2009)○ Detecting sadness in 140 characters.

●Jansen et al. (2009)○ Twitter power: Tweets as electronic word of mouth.

●Pandey et al. (2010)○ Sentiment analysis of microblogs.

● Volvoka et al. (2013)○ Exploring sentiment in social media: Bootstrapping subjectivity clues from

multilingual twitter streams

Only focused on positive/negative polarity classification

ASONAM 2015 Conference

ASONAM 2015 Conference

●o

●

●o

A multilingual system should incorporate new languages easily

ASONAM 2015 Conference

●o

ASONAM 2015 Conference

ASONAM 2015 Conference

●○ …

●○

Locked in at ASONAM Conference #happy

ASONAM 2015 Conference will be hosted in Paris this year

ASONAM 2015 Conference

●○○○○

ASONAM 2015 Conference

••

•• …

How to obtain them in a way that is independent of the language?

ASONAM 2015 Conference

●○

●○

ASONAM 2015 Conference

ASONAM 2015 Conference

•

•

ASONAM 2015 Conference

•

• emotion concepts will tend to be clustered together (as in many emotion words interconnected by the same connector words)

•

ASONAM 2015 Conference

PW

PW

CW

PW

ASONAM 2015 Conference

●•

•••

ASONAM 2015 Conference

•

ASONAM 2015 Conference

•••

……

••

…

… …

ASONAM 2015 Conference

•

••

…

… … …

ASONAM 2015 Conference

•

Anticipation:

p1 p2pn...

Joy:

p1 p2pn...

Trust:

p1 p2pn...

How do we calculate which patterns are more relavant to each emotion?

ASONAM 2015 Conference

•

•

••••

ASONAM 2015 Conference

•

•

ASONAM 2015 Conference

☺… …

☺… …

☺… …

☺

☺

ASONAM 2015 Conference

•

•

ASONAM 2015 Conference

… … … …

ASONAM 2015 Conference

•

•

ASONAM 2015 Conference

•

ASONAM 2015 Conference

ASONAM 2015 Conference

•

ASONAM 2015 Conference

ASONAM 2015 Conference