SlideShare uma empresa Scribd logo
1 de 40
Baixar para ler offline
University of Sheffield, NLP

Multimodal Sentiment Analysis of
Social Media
Diana Maynard
University of Sheffield, UK
David Dupplaw, Jonathon Hare
University of Southampton, UK
Twitter is full of mindless drivel
●

●
●

●
●

OMMMFG!!! JUST HEARD EMINEM'S “RAPGOD”. SMFH!!!
these other dudes might as well stop rapping if they not on this
level
i've got dressed but only because I need biscuits
I used to be so bad at naming any k idol group members pmsl I
would get so confused and now I'm pro ;)))
Pride and Prejudice is on. Oh my oh my be still my beating heart.
im gonna learn to be a lifegaurd hopfully so while everyone else is
working in a shop actually doing stuff il be sitting on a pool
side.yay
What are people reading
about?
●

●

Of the top 10 Twitter
accounts with the highest
number of followers:
●
7 pop stars
●
2 social media sites
●
and Barack Obama
Why on earth do we care
about this stuff?
We are the Wombles of Social Media!
Not because we like having forty winks...
But because we make good use of “things
that everyday folks leave behind”
There can be surprising value in trivia
●

Germtracker: deriving accurate real-time epidemiological
information from tweets [Sadilek, 2012]
●
Twitter analysis to predict who would get flu, and
to identify restaurants with a high risk of food
poisoning
●
Investigated the role of interactions between
users on social media on the real-life spread of
the disease
Even the mindless drivel could be useful
●

●

●

●

●

OMMMFG!!! JUST HEARD EMINEM'S “RAPGOD”. SMFH!!! these
other dudes might as well stop rapping if they not on this level
i've got dressed but only because I need biscuits
I used to be so bad at naming any k idol group members pmsl I
would get so confused and now I'm pro ;)))
Pride and Prejudice is on. Oh my oh my be still my beating heart.
im gonna learn to be a lifegaurd hopfully so while everyone else is
working in a shop actually doing stuff il be sitting on a pool side.yay
➔

English people like biscuits. A lot.

➔

What do young people think about their future careers?

➔

What do the public think about period dramas/films?

➔

People who like K Idol and RapGod also like Apple
products
Questions we want to answer
●

●

●
●
●

What are the opinions on crucial
social events and the key people
involved?
How are these opinions distributed
in relation to demographic user
data?
How have these opinions evolved?
Who are the opinion leaders?
What is their impact and influence?
Why?
●

●

●

Help archivists select material for inclusion in an archive of social
media for preserving community memories
Help journalists answer the “Big 5” questions: who, what, when,
where and why
Entities, events, topics and opinions can be used to categorise the
texts in different ways
●

●

Entities and events can be used as targets for opinion mining,
●

●

look at all texts about Obama (or Justin Bieber)
see what people thought about Obama (or Justin Bieber)

When linked to an ontology and/or combined with other information,
they can be used for reasoning about things not explicit in the text
●

see how opinions about different American presidents (or
Canadian pop stars) have changed over the years
Our approach
●

Combine opinion mining from text and multimedia

●

Rule-based approach to opinion mining from text
●
●

●

●

Can be adapted to the domain and language
Centred around entities, terms and events as the
opinion targets
Uses sentiment lexicons and linguistic subcomponents

Multimedia: determine sentiment from images
●
●

with and without specific context
use to support the text-based opinion mining
Basic approach to opinion finding in text
●

●

●

●

●

●
●

Run NER and term recognition using GATE tools: ANNIE/TwitIE +
TermRaider
Find sentiment-containing words via gazetteer lookup and allocate
sentiment strength scores
Find linguistic relations between sentiment words and entities/events
(opinion-target matching)
Use a number of linguistic sub-components to deal with issues such
as negatives, irony, swear words, sarcasm etc.
Modify sentiment strength scores based on the effect of these
modifiers (e.g. swear words and adverbs intensify score, negation
and sarcasm reverse the polarity)
Annotate entities/events and sentences with opinion scores
Aggregate scores as required, e.g. over all instances of an entity in
the document, over all sentences in the document.
Text-based opinion mining
Opinions on Greek Crisis
Challenges for NLP
●

●
●

●
●

Noisy language: unusual punctuation, capitalisation,
spelling, use of slang, sarcasm etc.
Terse nature of microposts such as tweets
Use of hashtags, @mentions etc causes problems for
tokenisation #thisistricky
Lack of context gives rise to ambiguities
NER performs poorly on microposts, mainly because of
linguistic pre-processing failure
●

●
●

Running standard IE tools (ANNIE) on 300 news articles –
87% F-measure
Running ANNIE on some tweets - < 40% F-measure
Other tools (e.g. Stanford NER) can reach even lower scores
Sarcasm is a part of British culture
●

So much so that the BBC has its own webpage on sarcasm
designed to teach non-native English speakers how to be
sarcastic successfully in conversation
My friend Barry likes Apple products
Or does he?
What does sarcasm do to polarity?
●

●

●

●

Sarcasm often indicated by hashtags in tweets such as
#sarcasm, #irony, #whoknew etc.
In general, when someone is being sarcastic, they're saying
the opposite of what they mean
So as long as you know which bit of the utterance is the
sarcastic bit, you can simply reverse the polarity
If there is no polarity on the original statement, you probably
just want to add a negative sentiment to it
●

●

It's not like I wanted to eat breakfast anyway
#sarcasm

To get the polarity scope right, you need to investigate the
hashtags: if there's more than one, you need to look at any
sentiment contained in them.
Identifying the scope of sarcasm
I am not happy that I woke up at 5:15 this
morning. #greatstart #sarcasm

You are really mature. #lying #sarcasm
Identifying the scope of sarcasm
I am not happy that I woke up at 5:15 this morning. #greatstart
#sarcasm
●
●

negative sentiment + positive hashtag + sarcasm hashtag
Chances are that the positive hashtag becomes negative with
sarcasm
You are really mature. #lying #sarcasm

●
●

●

positive sentiment + sarcasm hashtag + sarcasm hashtag
Chances are the positive sentiment is turned negative by both
sarcasm hashtags
When in doubt, it's usually safe to assume that a sarcastic
statement carries negative sentiment
Experiment with sarcastic hashtags




Collected a corpus of 134 tweets containing the hashtag
#sarcasm
Manually annotated sentences with sentiment








266 sentences, of which 68 opinionated (25%)
62 negative, 6 positive

Also annotated the same corpus as if the sarcasm was absent
Compared how well our applications performed on each, with
and without sarcasm analysis
The results were a little surprising
Effect of sarcasm on sentiment analysis
Sarcastic corpus

Recall

F1

Opinionated

74.58

63.77

68.75

Opinion+polarity - Regular

20.34

17.39

18.75

Polarity-only - Regular

27.27

27.27

27.27

Opinion+polarity - Sarcastic

57.63

49.28

53.13

Polarity-only - Sarcastic



Precision

77.02

77.28

77.28

Even when we knew the statement was sarcastic, we didn't
always get the polarity of the opinion right
Analysing Hashtags
What's in a hashtag?
●

Hashtags often contain smushed words
●
●
●

●

●

For NER we want the individual tokens
so we can link them to the right entity
For opinion mining, individual words in
the hashtags often indicate sentiment,
sarcasm etc.
●
●

●

#SteveJobs
#CombineAFoodAndABand
#southamerica

#greatidea
#worstdayever

We need to retokenise hashtags so that
we can use the content in our application
How to analyse hashtags?
●

●

●

●

Camelcasing makes it relatively easy to separate the
words, using an adapted tokeniser, but many people don't
bother
We use a simple approach based on dictionary matching
the longest consecutive strings, working L to R
●
#lifeisgreat -> #-life-is-great
●
#lovinglife -> #-loving-life
It's not foolproof, however
●
#greatstart -> #-greats-tart
In an experiment with non-camelcased English hashtags:
86.91% Precision, 90% Recall, F-measure 88.43%.
Introducing Images into Sentiment
Analysis
●

Three main ways to introduce multimedia
information:
●

●

●

Determine sentiment from an image in a specific
context (multimodal opinion mining)
Determine sentiment from an image in no
specific context (image opinion mining)
Support text sentiment analysis using
information from images (multimodal opinion
mining)
Multimodal Analysis
●

●

Using text as a means for reducing multimedia
search space to achieve contextual sentiment
analysis.
Cannot train computer
to recognise all people
in the world. Reduce
search space by using
entity extraction from
text and training person
recogniser on small set.
Multimodal Analysis

Steve Jobs Recogniser
Multimodal Analysis

Steve Jobs Recogniser
Multimodal Analysis
Face Model Fitting

Expression Recognition

Happiness
Image Sentiment Analysis

NEGATIVE

POSITIVE
Image Sentiment Analysis

Feature Correlation for
575,000 images from Flickr

NEGATIVE

POSITIVE
Image Sentiment Analysis
Conclusions
●

●

●
●

●
●

Novel approach to opinion mining from social media
combining text and multimedia processing
Both are really hard to do well, especially when your sources
are less than perfect
Often it's a chicken and egg problem
The fact that social media is full of rubbish doesn't make it
uninteresting
These new forms of data require significantly new strategies
We have interesting results already, but there's still a long way
to go
And for anyone who didn't grow up in the UK

●

Here are the Wombles
http://www.youtube.com/watch?v=aCf_PpDUTdA

Mais conteúdo relacionado

Mais procurados

Mais procurados (20)

Deep residual learning for image recognition
Deep residual learning for image recognitionDeep residual learning for image recognition
Deep residual learning for image recognition
 
Facial Expression Recognition (FER) using Deep Learning
Facial Expression Recognition (FER) using Deep LearningFacial Expression Recognition (FER) using Deep Learning
Facial Expression Recognition (FER) using Deep Learning
 
Yolo v2 ai_tech_20190421
Yolo v2 ai_tech_20190421Yolo v2 ai_tech_20190421
Yolo v2 ai_tech_20190421
 
Thomas Wolf "Transfer learning in NLP"
Thomas Wolf "Transfer learning in NLP"Thomas Wolf "Transfer learning in NLP"
Thomas Wolf "Transfer learning in NLP"
 
Generative Adversarial Networks and Their Applications in Medical Imaging
Generative Adversarial Networks  and Their Applications in Medical ImagingGenerative Adversarial Networks  and Their Applications in Medical Imaging
Generative Adversarial Networks and Their Applications in Medical Imaging
 
Michal Erel's SIFT presentation
Michal Erel's SIFT presentationMichal Erel's SIFT presentation
Michal Erel's SIFT presentation
 
[Paper] Multiscale Vision Transformers(MVit)
[Paper] Multiscale Vision Transformers(MVit)[Paper] Multiscale Vision Transformers(MVit)
[Paper] Multiscale Vision Transformers(MVit)
 
Noise
NoiseNoise
Noise
 
Contrast enhancement in digital images
Contrast enhancement in digital imagesContrast enhancement in digital images
Contrast enhancement in digital images
 
Histogram Specification or Matching Problem
Histogram Specification or Matching ProblemHistogram Specification or Matching Problem
Histogram Specification or Matching Problem
 
Sentiment analysis
Sentiment analysisSentiment analysis
Sentiment analysis
 
Lzw
LzwLzw
Lzw
 
Resnet
ResnetResnet
Resnet
 
Variational Autoencoders For Image Generation
Variational Autoencoders For Image GenerationVariational Autoencoders For Image Generation
Variational Autoencoders For Image Generation
 
Fill area algorithms
Fill area algorithmsFill area algorithms
Fill area algorithms
 
Intensity Transformation
Intensity TransformationIntensity Transformation
Intensity Transformation
 
Computer Vision: Correlation, Convolution, and Gradient
Computer Vision: Correlation, Convolution, and GradientComputer Vision: Correlation, Convolution, and Gradient
Computer Vision: Correlation, Convolution, and Gradient
 
Deep Learning - CNN and RNN
Deep Learning - CNN and RNNDeep Learning - CNN and RNN
Deep Learning - CNN and RNN
 
CVPR 2018 Paper Reading MobileNet V2
CVPR 2018 Paper Reading MobileNet V2CVPR 2018 Paper Reading MobileNet V2
CVPR 2018 Paper Reading MobileNet V2
 
Image restoration
Image restorationImage restoration
Image restoration
 

Destaque

Sentiment Analysis and Social Media: How and Why
Sentiment Analysis and Social Media: How and WhySentiment Analysis and Social Media: How and Why
Sentiment Analysis and Social Media: How and Why
Davide Feltoni Gurini
 

Destaque (20)

How Sentiment Analysis works
How Sentiment Analysis worksHow Sentiment Analysis works
How Sentiment Analysis works
 
Introduction to Sentiment Analysis
Introduction to Sentiment AnalysisIntroduction to Sentiment Analysis
Introduction to Sentiment Analysis
 
Tiffany Iwantoro_Exchange Rate Directional Forecasting using Sentiment Analys...
Tiffany Iwantoro_Exchange Rate Directional Forecasting using Sentiment Analys...Tiffany Iwantoro_Exchange Rate Directional Forecasting using Sentiment Analys...
Tiffany Iwantoro_Exchange Rate Directional Forecasting using Sentiment Analys...
 
Science Festival 2011 Programme
Science Festival 2011 ProgrammeScience Festival 2011 Programme
Science Festival 2011 Programme
 
Knime customer intelligence on social media: Text Analytics vs. Network Mining
Knime customer intelligence on social media: Text Analytics vs. Network MiningKnime customer intelligence on social media: Text Analytics vs. Network Mining
Knime customer intelligence on social media: Text Analytics vs. Network Mining
 
Social Data Sentiment Analysis
Social Data Sentiment AnalysisSocial Data Sentiment Analysis
Social Data Sentiment Analysis
 
A review of sentiment analysis approaches in big
A review of sentiment analysis approaches in bigA review of sentiment analysis approaches in big
A review of sentiment analysis approaches in big
 
Scalable sentiment classification for big data analysis using naive bayes cla...
Scalable sentiment classification for big data analysis using naive bayes cla...Scalable sentiment classification for big data analysis using naive bayes cla...
Scalable sentiment classification for big data analysis using naive bayes cla...
 
Challenges of social media analysis in the real world
Challenges of social media analysis in the real worldChallenges of social media analysis in the real world
Challenges of social media analysis in the real world
 
Social media & sentiment analysis splunk conf2012
Social media & sentiment analysis   splunk conf2012Social media & sentiment analysis   splunk conf2012
Social media & sentiment analysis splunk conf2012
 
Sentiment Analysis of Film-Related Messages on Social Media
Sentiment Analysis of Film-Related Messages on Social MediaSentiment Analysis of Film-Related Messages on Social Media
Sentiment Analysis of Film-Related Messages on Social Media
 
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...
 
Text Classification, Sentiment Analysis, and Opinion Mining
Text Classification, Sentiment Analysis, and Opinion MiningText Classification, Sentiment Analysis, and Opinion Mining
Text Classification, Sentiment Analysis, and Opinion Mining
 
Opinion mining for social media
Opinion mining for social mediaOpinion mining for social media
Opinion mining for social media
 
Multimodal man machine interaction
Multimodal man machine interactionMultimodal man machine interaction
Multimodal man machine interaction
 
Applying large scale text analytics with graph databases
Applying large scale text analytics with graph databasesApplying large scale text analytics with graph databases
Applying large scale text analytics with graph databases
 
Practical sentiment analysis
Practical sentiment analysisPractical sentiment analysis
Practical sentiment analysis
 
Sentiment Analysis and Social Media: How and Why
Sentiment Analysis and Social Media: How and WhySentiment Analysis and Social Media: How and Why
Sentiment Analysis and Social Media: How and Why
 
Sentiment analysis using naive bayes classifier
Sentiment analysis using naive bayes classifier Sentiment analysis using naive bayes classifier
Sentiment analysis using naive bayes classifier
 
Big Data & Text Mining
Big Data & Text MiningBig Data & Text Mining
Big Data & Text Mining
 

Semelhante a Multimodal opinion mining from social media

Can Social Media Analysis Improve Collective Awareness of Climate Change?
Can Social Media Analysis Improve Collective Awareness of Climate Change?Can Social Media Analysis Improve Collective Awareness of Climate Change?
Can Social Media Analysis Improve Collective Awareness of Climate Change?
Diana Maynard
 
2. Research-2.pptx
2. Research-2.pptx2. Research-2.pptx
2. Research-2.pptx
CJ Stevens
 

Semelhante a Multimodal opinion mining from social media (20)

Do we really know what people mean when they tweet?
Do we really know what people mean when they tweet?Do we really know what people mean when they tweet?
Do we really know what people mean when they tweet?
 
Can Social Media Analysis Improve Collective Awareness of Climate Change?
Can Social Media Analysis Improve Collective Awareness of Climate Change?Can Social Media Analysis Improve Collective Awareness of Climate Change?
Can Social Media Analysis Improve Collective Awareness of Climate Change?
 
Who cares about sarcastic tweets? Investigating the impact of sarcasm on sent...
Who cares about sarcastic tweets? Investigating the impact of sarcasm on sent...Who cares about sarcastic tweets? Investigating the impact of sarcasm on sent...
Who cares about sarcastic tweets? Investigating the impact of sarcasm on sent...
 
Can Social Media Analysis Improve Collective Awareness of Climate Change?
Can Social Media Analysis Improve Collective Awareness of Climate Change?Can Social Media Analysis Improve Collective Awareness of Climate Change?
Can Social Media Analysis Improve Collective Awareness of Climate Change?
 
Cls8 decarbonet
Cls8 decarbonetCls8 decarbonet
Cls8 decarbonet
 
Listen. Engage. Build Relationships. Spring 2014
Listen. Engage. Build Relationships. Spring 2014Listen. Engage. Build Relationships. Spring 2014
Listen. Engage. Build Relationships. Spring 2014
 
Social media for the voice actor
Social media for the voice actorSocial media for the voice actor
Social media for the voice actor
 
Turning findings into insights
Turning findings into insightsTurning findings into insights
Turning findings into insights
 
How to increase your instagram engagement (cm)
How to increase your instagram engagement (cm)How to increase your instagram engagement (cm)
How to increase your instagram engagement (cm)
 
Social Media: Friend or Foe (May 2014)
Social Media: Friend or Foe (May 2014)Social Media: Friend or Foe (May 2014)
Social Media: Friend or Foe (May 2014)
 
Ringling College of Art & Design: Content and Social Media
Ringling College of Art & Design: Content and Social MediaRingling College of Art & Design: Content and Social Media
Ringling College of Art & Design: Content and Social Media
 
Using Social Media as a Project Management Professional
Using Social Media as a Project Management ProfessionalUsing Social Media as a Project Management Professional
Using Social Media as a Project Management Professional
 
What do you really mean when you tweet? Challenges for opinion mining on soci...
What do you really mean when you tweet? Challenges for opinion mining on soci...What do you really mean when you tweet? Challenges for opinion mining on soci...
What do you really mean when you tweet? Challenges for opinion mining on soci...
 
2016 datascience emotion analysis - english version
2016 datascience emotion analysis - english version2016 datascience emotion analysis - english version
2016 datascience emotion analysis - english version
 
Made in Millersville Social Media Planbook
Made in Millersville Social Media PlanbookMade in Millersville Social Media Planbook
Made in Millersville Social Media Planbook
 
RED BOAT - Online Learning Notes (July 2021)
RED BOAT - Online Learning Notes (July 2021)RED BOAT - Online Learning Notes (July 2021)
RED BOAT - Online Learning Notes (July 2021)
 
doc
docdoc
doc
 
2. Research.pptx
2. Research.pptx2. Research.pptx
2. Research.pptx
 
2. Research-2.pptx
2. Research-2.pptx2. Research-2.pptx
2. Research-2.pptx
 
Aardman Social Media Training Slides
Aardman Social Media Training SlidesAardman Social Media Training Slides
Aardman Social Media Training Slides
 

Mais de Diana Maynard

Mais de Diana Maynard (19)

Filth and lies: analysing social media
Filth and lies: analysing social mediaFilth and lies: analysing social media
Filth and lies: analysing social media
 
Adding value to NLP: a little semantics goes a long way
Adding value to NLP: a little semantics goes a long wayAdding value to NLP: a little semantics goes a long way
Adding value to NLP: a little semantics goes a long way
 
Methodological possibilities for strengthening the monitoring of SDG indicato...
Methodological possibilities for strengthening the monitoring of SDG indicato...Methodological possibilities for strengthening the monitoring of SDG indicato...
Methodological possibilities for strengthening the monitoring of SDG indicato...
 
Getting the-most-out-of-conferences
Getting the-most-out-of-conferencesGetting the-most-out-of-conferences
Getting the-most-out-of-conferences
 
Understanding the world with NLP: interactions between society, behaviour and...
Understanding the world with NLP: interactions between society, behaviour and...Understanding the world with NLP: interactions between society, behaviour and...
Understanding the world with NLP: interactions between society, behaviour and...
 
The language of social media
The language of social mediaThe language of social media
The language of social media
 
Text analysis-semantic-search
Text analysis-semantic-searchText analysis-semantic-search
Text analysis-semantic-search
 
Using language to save the world: interactions between society, behaviour and...
Using language to save the world: interactions between society, behaviour and...Using language to save the world: interactions between society, behaviour and...
Using language to save the world: interactions between society, behaviour and...
 
Adapting NLP tools to diverse data: challenges and solutions
Adapting NLP tools to diverse data: challenges and solutionsAdapting NLP tools to diverse data: challenges and solutions
Adapting NLP tools to diverse data: challenges and solutions
 
Ontologies as bridges between data sources and user queries: the KNOWMAK proj...
Ontologies as bridges between data sources and user queries: the KNOWMAK proj...Ontologies as bridges between data sources and user queries: the KNOWMAK proj...
Ontologies as bridges between data sources and user queries: the KNOWMAK proj...
 
20 Years of Text Mining Applications with GATE: from Donald Trump to curing c...
20 Years of Text Mining Applications with GATE: from Donald Trump to curing c...20 Years of Text Mining Applications with GATE: from Donald Trump to curing c...
20 Years of Text Mining Applications with GATE: from Donald Trump to curing c...
 
Text Analysis and Semantic Search with GATE
Text Analysis and Semantic Search with GATEText Analysis and Semantic Search with GATE
Text Analysis and Semantic Search with GATE
 
GATE: a text analysis tool for social media
GATE: a text analysis tool for social mediaGATE: a text analysis tool for social media
GATE: a text analysis tool for social media
 
Tools for (Almost) Real-Time Social Media Analysis
Tools for (Almost) Real-Time Social Media AnalysisTools for (Almost) Real-Time Social Media Analysis
Tools for (Almost) Real-Time Social Media Analysis
 
Social media analytics as a service: tools from GATE
Social media analytics as a service: tools from GATESocial media analytics as a service: tools from GATE
Social media analytics as a service: tools from GATE
 
Text analysis and Semantic Search with GATE
Text analysis and Semantic Search with GATEText analysis and Semantic Search with GATE
Text analysis and Semantic Search with GATE
 
Disability and Adventure Travel: the Double-Edged Sword
Disability and Adventure Travel: the Double-Edged SwordDisability and Adventure Travel: the Double-Edged Sword
Disability and Adventure Travel: the Double-Edged Sword
 
Practical Opinion Mining for Social Media
Practical Opinion Mining for Social MediaPractical Opinion Mining for Social Media
Practical Opinion Mining for Social Media
 
A tailor-made one-size-fits-all approach to sentiment analysis
A tailor-made one-size-fits-all approach to sentiment analysisA tailor-made one-size-fits-all approach to sentiment analysis
A tailor-made one-size-fits-all approach to sentiment analysis
 

Último

Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
Joaquim Jorge
 

Último (20)

Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin WoodPolkadot JAM Slides - Token2049 - By Dr. Gavin Wood
Polkadot JAM Slides - Token2049 - By Dr. Gavin Wood
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
 
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
 
Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024Partners Life - Insurer Innovation Award 2024
Partners Life - Insurer Innovation Award 2024
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot TakeoffStrategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
Apidays New York 2024 - The Good, the Bad and the Governed by David O'Neill, ...
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)
 
Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024Manulife - Insurer Innovation Award 2024
Manulife - Insurer Innovation Award 2024
 
MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024
 
HTML Injection Attacks: Impact and Mitigation Strategies
HTML Injection Attacks: Impact and Mitigation StrategiesHTML Injection Attacks: Impact and Mitigation Strategies
HTML Injection Attacks: Impact and Mitigation Strategies
 

Multimodal opinion mining from social media

  • 1. University of Sheffield, NLP Multimodal Sentiment Analysis of Social Media Diana Maynard University of Sheffield, UK David Dupplaw, Jonathon Hare University of Southampton, UK
  • 2. Twitter is full of mindless drivel ● ● ● ● ● OMMMFG!!! JUST HEARD EMINEM'S “RAPGOD”. SMFH!!! these other dudes might as well stop rapping if they not on this level i've got dressed but only because I need biscuits I used to be so bad at naming any k idol group members pmsl I would get so confused and now I'm pro ;))) Pride and Prejudice is on. Oh my oh my be still my beating heart. im gonna learn to be a lifegaurd hopfully so while everyone else is working in a shop actually doing stuff il be sitting on a pool side.yay
  • 3. What are people reading about? ● ● Of the top 10 Twitter accounts with the highest number of followers: ● 7 pop stars ● 2 social media sites ● and Barack Obama Why on earth do we care about this stuff?
  • 4. We are the Wombles of Social Media!
  • 5. Not because we like having forty winks...
  • 6. But because we make good use of “things that everyday folks leave behind”
  • 7.
  • 8. There can be surprising value in trivia ● Germtracker: deriving accurate real-time epidemiological information from tweets [Sadilek, 2012] ● Twitter analysis to predict who would get flu, and to identify restaurants with a high risk of food poisoning ● Investigated the role of interactions between users on social media on the real-life spread of the disease
  • 9.
  • 10. Even the mindless drivel could be useful ● ● ● ● ● OMMMFG!!! JUST HEARD EMINEM'S “RAPGOD”. SMFH!!! these other dudes might as well stop rapping if they not on this level i've got dressed but only because I need biscuits I used to be so bad at naming any k idol group members pmsl I would get so confused and now I'm pro ;))) Pride and Prejudice is on. Oh my oh my be still my beating heart. im gonna learn to be a lifegaurd hopfully so while everyone else is working in a shop actually doing stuff il be sitting on a pool side.yay ➔ English people like biscuits. A lot. ➔ What do young people think about their future careers? ➔ What do the public think about period dramas/films? ➔ People who like K Idol and RapGod also like Apple products
  • 11. Questions we want to answer ● ● ● ● ● What are the opinions on crucial social events and the key people involved? How are these opinions distributed in relation to demographic user data? How have these opinions evolved? Who are the opinion leaders? What is their impact and influence?
  • 12. Why? ● ● ● Help archivists select material for inclusion in an archive of social media for preserving community memories Help journalists answer the “Big 5” questions: who, what, when, where and why Entities, events, topics and opinions can be used to categorise the texts in different ways ● ● Entities and events can be used as targets for opinion mining, ● ● look at all texts about Obama (or Justin Bieber) see what people thought about Obama (or Justin Bieber) When linked to an ontology and/or combined with other information, they can be used for reasoning about things not explicit in the text ● see how opinions about different American presidents (or Canadian pop stars) have changed over the years
  • 13.
  • 14. Our approach ● Combine opinion mining from text and multimedia ● Rule-based approach to opinion mining from text ● ● ● ● Can be adapted to the domain and language Centred around entities, terms and events as the opinion targets Uses sentiment lexicons and linguistic subcomponents Multimedia: determine sentiment from images ● ● with and without specific context use to support the text-based opinion mining
  • 15. Basic approach to opinion finding in text ● ● ● ● ● ● ● Run NER and term recognition using GATE tools: ANNIE/TwitIE + TermRaider Find sentiment-containing words via gazetteer lookup and allocate sentiment strength scores Find linguistic relations between sentiment words and entities/events (opinion-target matching) Use a number of linguistic sub-components to deal with issues such as negatives, irony, swear words, sarcasm etc. Modify sentiment strength scores based on the effect of these modifiers (e.g. swear words and adverbs intensify score, negation and sarcasm reverse the polarity) Annotate entities/events and sentences with opinion scores Aggregate scores as required, e.g. over all instances of an entity in the document, over all sentences in the document.
  • 18.
  • 19. Challenges for NLP ● ● ● ● ● Noisy language: unusual punctuation, capitalisation, spelling, use of slang, sarcasm etc. Terse nature of microposts such as tweets Use of hashtags, @mentions etc causes problems for tokenisation #thisistricky Lack of context gives rise to ambiguities NER performs poorly on microposts, mainly because of linguistic pre-processing failure ● ● ● Running standard IE tools (ANNIE) on 300 news articles – 87% F-measure Running ANNIE on some tweets - < 40% F-measure Other tools (e.g. Stanford NER) can reach even lower scores
  • 20. Sarcasm is a part of British culture ● So much so that the BBC has its own webpage on sarcasm designed to teach non-native English speakers how to be sarcastic successfully in conversation
  • 21. My friend Barry likes Apple products
  • 23. What does sarcasm do to polarity? ● ● ● ● Sarcasm often indicated by hashtags in tweets such as #sarcasm, #irony, #whoknew etc. In general, when someone is being sarcastic, they're saying the opposite of what they mean So as long as you know which bit of the utterance is the sarcastic bit, you can simply reverse the polarity If there is no polarity on the original statement, you probably just want to add a negative sentiment to it ● ● It's not like I wanted to eat breakfast anyway #sarcasm To get the polarity scope right, you need to investigate the hashtags: if there's more than one, you need to look at any sentiment contained in them.
  • 24. Identifying the scope of sarcasm I am not happy that I woke up at 5:15 this morning. #greatstart #sarcasm You are really mature. #lying #sarcasm
  • 25. Identifying the scope of sarcasm I am not happy that I woke up at 5:15 this morning. #greatstart #sarcasm ● ● negative sentiment + positive hashtag + sarcasm hashtag Chances are that the positive hashtag becomes negative with sarcasm You are really mature. #lying #sarcasm ● ● ● positive sentiment + sarcasm hashtag + sarcasm hashtag Chances are the positive sentiment is turned negative by both sarcasm hashtags When in doubt, it's usually safe to assume that a sarcastic statement carries negative sentiment
  • 26. Experiment with sarcastic hashtags   Collected a corpus of 134 tweets containing the hashtag #sarcasm Manually annotated sentences with sentiment      266 sentences, of which 68 opinionated (25%) 62 negative, 6 positive Also annotated the same corpus as if the sarcasm was absent Compared how well our applications performed on each, with and without sarcasm analysis The results were a little surprising
  • 27. Effect of sarcasm on sentiment analysis Sarcastic corpus Recall F1 Opinionated 74.58 63.77 68.75 Opinion+polarity - Regular 20.34 17.39 18.75 Polarity-only - Regular 27.27 27.27 27.27 Opinion+polarity - Sarcastic 57.63 49.28 53.13 Polarity-only - Sarcastic  Precision 77.02 77.28 77.28 Even when we knew the statement was sarcastic, we didn't always get the polarity of the opinion right
  • 29. What's in a hashtag? ● Hashtags often contain smushed words ● ● ● ● ● For NER we want the individual tokens so we can link them to the right entity For opinion mining, individual words in the hashtags often indicate sentiment, sarcasm etc. ● ● ● #SteveJobs #CombineAFoodAndABand #southamerica #greatidea #worstdayever We need to retokenise hashtags so that we can use the content in our application
  • 30. How to analyse hashtags? ● ● ● ● Camelcasing makes it relatively easy to separate the words, using an adapted tokeniser, but many people don't bother We use a simple approach based on dictionary matching the longest consecutive strings, working L to R ● #lifeisgreat -> #-life-is-great ● #lovinglife -> #-loving-life It's not foolproof, however ● #greatstart -> #-greats-tart In an experiment with non-camelcased English hashtags: 86.91% Precision, 90% Recall, F-measure 88.43%.
  • 31. Introducing Images into Sentiment Analysis ● Three main ways to introduce multimedia information: ● ● ● Determine sentiment from an image in a specific context (multimodal opinion mining) Determine sentiment from an image in no specific context (image opinion mining) Support text sentiment analysis using information from images (multimodal opinion mining)
  • 32. Multimodal Analysis ● ● Using text as a means for reducing multimedia search space to achieve contextual sentiment analysis. Cannot train computer to recognise all people in the world. Reduce search space by using entity extraction from text and training person recogniser on small set.
  • 35. Multimodal Analysis Face Model Fitting Expression Recognition Happiness
  • 37. Image Sentiment Analysis Feature Correlation for 575,000 images from Flickr NEGATIVE POSITIVE
  • 39. Conclusions ● ● ● ● ● ● Novel approach to opinion mining from social media combining text and multimedia processing Both are really hard to do well, especially when your sources are less than perfect Often it's a chicken and egg problem The fact that social media is full of rubbish doesn't make it uninteresting These new forms of data require significantly new strategies We have interesting results already, but there's still a long way to go
  • 40. And for anyone who didn't grow up in the UK ● Here are the Wombles http://www.youtube.com/watch?v=aCf_PpDUTdA