Sentiment, Opinion & Emotion on the Multilingual Web
1. Sentiment, Opinion &
Emotion on the
Multilingual Web
Seth Grimes
Alta Plana Corporation
@sethgrimes
New Horizons for the
Multilingal Web – Madrid
May 8, 2014
3. Sentiment, Opinion & Emotion
Four varieties of data
1. Machine data (e.g., logs, sensor outputs,
clickstreams).
2. Interactions and transactions (including
location and time).
3. Profile: individual, demographic & behavioral.
4. Media: Text, audio, images, and video.
And two super-types
• Facts.
• Feelings.
4. Sentiment, Opinion & Emotion
Feelings
“Sentiment analysis or opinion mining is the
computational study of opinions, sentiments and
emotions expressed in text.”
-- Bing Liu, 2010, “Sentiment Analysis and
Subjectivity,” in Handbook of Natural Language
Processing
• “My graphics look horrible when I play and I’m all
laggy.”
• “The Iraqi regime… possesses and produces chemical
and biological weapons.” -- George W. Bush, Oct 7,
2002.
• “I like pepsi better simple as that, coke feels like I’m
drinking battery acid but it tastes good.”
6. Sentiment, Opinion & Emotion
The sentiment value of an opinion may be
expressed as a quintuple (oj, fjk, soijkl, hi, tl)
where:
• oj is a target object.
• fjk is an feature of the object oj.
• hi is an opinion holder.
• tl is the time when the opinion is expressed.
• soijkl is the sentiment value of the opinion of
the opinion holder hi regarding feature fjk of
object oj at time tl.
• soijkl is +ve, -ve, or neu, or a more granular
rating.
(Bing Liu, NLP Handbook)
7. Sentiment, Opinion & Emotion
Comparative opinions --
(O1, O2, F, po, h, t):
• O1 and O2 are object sets being compared
based on shared features F.
• po is the preferred object set of the opinion
holder h.
• t is the time when the comparative opinion is
expressed.
(Bing Liu)
8. Sentiment, Opinion & Emotion
Questions for business (& government):
What are people saying? What’s hot/trending?
What are they saying about {topic|person|product} X?
... about X versus {topic|person|product} Y?
How has opinion about X and Y evolved?
How has opinion correlated with
{our|competitors’|general}
{news|marketing|sales|events}?
Who (and What, When & How) are opinion leaders?
How does sentiment propagate across multiple channels?
What’s behind opinion, the root causes?
(How) Can we link opinions, profiles, behaviors &
transactions to discern intent and predict actions?
(oj, fjk, soijkl, hi, tl)? (O1, O2, F, po, h, t)?
9. Sentiment, Opinion & Emotion
Sentiment
surfaced via
typical industry
applications.
11. Sentiment, Opinion & Emotion
Sentiment is of interest at multiple levels.
Corpus / data space, i.e., across multiple sources.
Document.
Statement / sentence.
Entity / topic / concept.
Human language is noisy and chaotic!
Jargon, slang, irony, ambiguity, anaphora,
polysemy, synonymy, etc., and culturally
dependent.
Context is key. Discourse analysis comes into
play.
Meaning may be implied rather than directly
stated.
15. Sentiment, Opinion & Emotion
Prediction/Feeling/Wish... and Intent
http://www.aiaioo.com/whitepapers/intention_analysis_use_cases.pdf
http://sentibet.com/
16. Sentiment, Opinion & Emotion
“It was the best of times, it was the worst of
times.” Opinion++?
Maybe (oj, fjk, hi, il,cm, mn, soijklmn) where:
• oj is a target object.
• fjk is an feature of the object oj.
• hi is an opinion holder.
• il is an instance of opinion expression.
• cm is a categorization.
• mn is a classification method.
• soijklmn is the sentiment value of the opinion of
opinion holder hi at time tl regarding feature
fjk of object oj, within categorization cm and as
classified via method mn .
18. Sentiment, Opinion & Emotion
All this keeps us
busy, even
without…
“The share rise in users who
selected
Arabic…coincided with
much of the civil unrest…
in Middle Eastern
countries.”
http://bits.blogs.nytimes.com/2014/03/09/t
he-languages-of-twitter-users/
21. Sentiment, Opinion &
Emotion on the
Multilingual Web
Seth Grimes
Alta Plana Corporation
@sethgrimes
New Horizons for the
Multilingal Web – Madrid
May 8, 2014