A Practical Guide to Sentiment Analysis

Visit the link below to download the full version of this book:

https://medipdf.com/product/a-practical-guide-to-sentiment-analysis/

Click Download Now

This exciting Book Series aims to publish state-of-the-art research on socially
intelligent, affective and multimodal human-machine interaction and systems.
It will emphasize the role of affect in social interactions and the humanistic side
of affective computing by promoting publications at the cross-roads between
engineering and human sciences (including biological, social and cultural aspects
of human life). Three broad domains of social and affective computing will be
covered by the book series: (1) social computing, (2) affective computing, and
(3) interplay of the first two domains (for example, augmenting social interaction
through affective computing). Examples of the first domain will include but not
limited to: all types of social interactions that contribute to the meaning, interest and
richness of our daily life, for example, information produced by a group of people
used to provide or enhance the functioning of a system. Examples of the second
domain will include, but not limited to: computational and psychological models of
emotions, bodily manifestations of affect (facial expressions, posture, behavior,
physiology), and affective interfaces and applications (dialogue systems, games,
learning etc.). This series will publish works of the highest quality that advance
the understanding and practical application of social and affective computing
techniques. Research monographs, introductory and advanced level textbooks,
volume editions and proceedings will be considered.

More information about this series at http://www.springer.com/series/13199

Erik Cambria • Dipankar Das
Sivaji Bandyopadhyay • Antonio Feraco
Editors

A Practical Guide to
Sentiment Analysis

123
Editors
Erik Cambria Dipankar Das
School of Computer Science Computer Science
and Engineering and Engineering Department
Nanyang Technological University Jadavpur University
Singapore, Singapore Kolkata, India

Sivaji Bandyopadhyay Antonio Feraco

Computer Science Fraunhofer IDM@NTU
and Engineering Department Nanyang Technological University
Jadavpur University Singapore, Singapore
Kolkata, India

ISSN 2509-5706 ISSN 2509-5714 (electronic)

Socio-Affective Computing
ISBN 978-3-319-55392-4 ISBN 978-3-319-55394-8 (eBook)
DOI 10.1007/978-3-319-55394-8

Library of Congress Control Number: 2017938021

© Springer International Publishing AG 2017

Chapter 4 is published with kind permission of the Her Majesty the Queen Right of Canada.
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

While sentiment analysis research has become very popular in the past ten years,
most companies and researchers still approach it simply as a polarity detection
problem. In reality, sentiment analysis is a “suitcase problem” that requires tackling
many natural language processing (NLP) subtasks, including microtext analysis,
sarcasm detection, anaphora resolution, subjectivity detection, and aspect extrac-
tion. In this book, we propose an overview of the main issues and challenges
associated with current sentiment analysis research and provide some insights on
practical tools and techniques that can be exploited to both advance the state of the
art in all sentiment analysis subtasks and explore new areas in the same context.
In Chap. 1, we discuss the state of the art of affective computing and sentiment
analysis research, including recent deep learning techniques and linguistic patterns
for emotion and polarity detection from different modalities, e.g., text and video.
In Chap. 2, Bing Liu describes different aspects of sentiment analysis and
different types of opinions. In particular, he uses product reviews as examples to
introduce general key concepts and definitions that are applicable to all forms of
formal and informal opinion text and all kinds of domains including social and
political domains.
In Chap. 3, Jiwei Li and Eduard Hovy describe possible directions for deeper
understanding about what opinions or sentiments are, why people hold them, and
why and how their facets are chosen and expressed, helping bridge the gap between
psychology/cognitive science and computational approaches.
In Chap. 4, Saif Mohammad discusses different sentiment analysis problems and
the challenges that are to be faced in order to go beyond simply determining whether
a piece of text is positive, negative, or neutral. In particular, the chapter aims to equip
researchers and practitioners with pointers to the latest developments in sentiment
analysis and encourage more work in the diverse landscape of problems, especially
those areas that are relatively less explored.
In Chap. 5, Aditya Joshi, Pushpak Bhattacharyya, and Sagar Ahire contrast the
process of lexicon creation for a new language or a resource-scarce language from
a resource-rich one and, hence, show how the produced sentiment resources can be
exploited to solve classic sentiment analysis problems.

v
vi Preface

In Chap. 6, Hongning Wang and ChengXiang Zhai show how generative models
can be used to integrate opinionated text data and their companion numerical
sentiment ratings, enabling deeper analysis of sentiment and opinions to obtain not
only subtopic-level sentiment but also latent relative weights on different subtopics.
In Chap. 7, Vasudeva Varma, Litton Kurisinkel, and Priya Radhakrishnan present
an overview of general approaches to automated text summarization with more
emphasis on extractive summarization techniques. They also describe recent works
on extractive summarization and the nature of scoring function for candidate
summary.
In Chap. 8, Paolo Rosso and Leticia Cagnina describe the very challenging
problems of deception detection and opinion spam detection, as lies and spam are
becoming increasingly serious issues with the rise, both in size and importance, of
social media and public opinion.
Finally, in Chap. 9 Federica Bisio et al. describe how to enhance the accuracy
of any algorithm for emotion or polarity detection through the integration of
commonsense reasoning resources, e.g., by embedding a concept-level knowledge
base for sentiment analysis.

Singapore, Singapore Erik Cambria

Kolkata, India Dipankar Das
Kolkata, India Sivaji Bandyopadhyay
Singapore, Singapore Antonio Feraco
Contents

1 Affective Computing and Sentiment Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Erik Cambria, Dipankar Das, Sivaji Bandyopadhyay,
and Antonio Feraco
2 Many Facets of Sentiment Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Bing Liu
3 Reflections on Sentiment/Opinion Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Jiwei Li and Eduard Hovy
4 Challenges in Sentiment Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Saif M. Mohammad
5 Sentiment Resources: Lexicons and Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Aditya Joshi, Pushpak Bhattacharyya, and Sagar Ahire
6 Generative Models for Sentiment Analysis and Opinion Mining. . . . . . . 107
Hongning Wang and ChengXiang Zhai
7 Social Media Summarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Vasudeva Varma, Litton J. Kurisinkel, and Priya Radhakrishnan
8 Deception Detection and Opinion Spam. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Paolo Rosso and Leticia C. Cagnina
9 Concept-Level Sentiment Analysis with SenticNet . . . . . . . . . . . . . . . . . . . . . . . 173
Federica Bisio, Claudia Meda, Paolo Gastaldo, Rodolfo Zunino,
and Erik Cambria

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

vii
Chapter 1
Affective Computing and Sentiment Analysis

Erik Cambria, Dipankar Das, Sivaji Bandyopadhyay, and Antonio Feraco

Abstract Understanding emotions is one of the most important aspects of personal

development and growth and, as such, it is a key tile for the emulation of
human intelligence. Besides being a important for the advancement of AI, emotion
processing is also important for the closely related task of polarity detection. The
opportunity automatically to capture the sentiments of the general public about
social events, political movements, marketing campaigns, and product preferences,
in fact, has raised increasing interest both in the scientific community, for the
exciting open challenges, and in the business world, for the remarkable fallouts
in marketing and financial market prediction. This has led to the emerging fields
of affective computing and sentiment analysis, which leverage on human-computer
interaction, information retrieval, and multimodal signal processing for distilling
people’s sentiments from the ever-growing amount of online social data.

Keywords Affective computing • Sentiment analysis • Five eras of the Web •

Jumping NLP curves • Hybrid approaches

1.1 Introduction

Emotions play an important role in successful and effective human-human relation-

ships. In fact, in many situations, human ‘emotional intelligence’ is more important
than IQ for successful interaction (Pantic et al. 2005). There is also significant
evidence that rational learning in humans is dependent on emotions (Picard 1997).

E. Cambria ()
School of Computer Science and Engineering, Nanyang Technological University, 639798,
Singapore, Singapore
e-mail: [email protected]
D. Das • S. Bandyopadhyay
Computer Science and Engineering Department, Jadavpur University, 700032, Kolkata, India
e-mail: [email protected]; [email protected]
A. Feraco
Fraunhofer IDM@NTU, Nanyang Technological University, Singapore, Singapore
e-mail: [email protected]

© Springer International Publishing AG 2017 1

E. Cambria et al. (eds.), A Practical Guide to Sentiment Analysis,
Socio-Affective Computing 5, DOI 10.1007/978-3-319-55394-8_1
2 E. Cambria et al.

Affective computing and sentiment analysis, hence, are key for the advancement
of AI (Minsky 2006) and all the research fields that stem from it. Moreover, they find
applications in several different scenarios and there is a good number of companies,
large and small, that include the analysis of emotions and sentiments as part of
their mission. Sentiment mining techniques can be exploited for the creation and
automated upkeep of review and opinion aggregation websites, in which opinionated
text and videos are continuously gathered from the Web and not restricted to
just product reviews, but also to wider topics such as political issues and brand
perception.
Affective computing and sentiment analysis have also a great potential as a
sub-component technology for other systems. They can enhance the capabilities
of customer relationship management and recommendation systems allowing, for
example, to find out which features customers are particularly happy about or
to exclude from the recommendations items that have received very negative
feedbacks. Similarly, they can be exploited for affective tutoring and affective enter-
tainment or for troll filtering and spam detection in online social communication.
Business intelligence is also one of the main factors behind corporate interest
in the fields of affective computing and sentiment analysis. Nowadays, companies
invest an increasing amount of money in marketing strategies and they are constantly
interested in both collecting and predicting the attitudes of the general public
towards their products and brands. The design of automatic tools capable to mine
sentiments over the Web in real-time and to create condensed versions of these
represents one of the most active research and development areas. The development
of such systems, moreover, is not only important for commercial purposes, but
also for government intelligence applications able to monitor increases in hostile
communications or to model cyber-issue diffusion.
Several commercial and academic tools, e.g., IBM,1 SAS,2 Oracle,3 SenticNet4
and Luminoso,5 track public viewpoints on a large-scale by offering graphical
summarizations of trends and opinions in the blogosphere. Nevertheless, most
commercial off-the-shelf (COTS) tools are limited to a polarity evaluation or a mood
classification according to a very limited set of emotions. In addition, such methods
mainly rely on parts of text in which emotional states are explicitly expressed
and, hence, they are unable to capture opinions and sentiments that are expressed
implicitly. Because they are mainly based on statistical properties associated with
words, in fact, many COTS tools are easily tricked by linguistic operators such as
negation and disjunction.
The remainder of this chapter lists common tasks of affective computing and
sentiment analysis and presents a general categorization for them, after which some
concluding remarks are proposed.

1
http://ibm.com/analytics
2
http://sas.com/social
3
http://oracle.com/social
4
http://business.sentic.net
5
http://luminoso.com
1 Affective Computing and Sentiment Analysis 3

1.2 Common Tasks

The Web is evolving towards an era where communities will define future products
and services.6 In this context, big social data analysis (Cambria et al. 2014) is
destined to attract increasing interest from both academia and business (Fig. 1.1).

Fig. 1.1 Owyang’s Five-Eras vision shows that mining sentiments from the general public is
becoming increasingly important for the future of the Web

6
http://web-strategist.com/blog/2009/04/27
4 E. Cambria et al.

The basic tasks of affective computing and sentiment analysis are emotion
recognition (Picard 1997; Calvo and D’Mello 2010; Zeng et al. 2009; Schuller et al.
2011; Gunes and Schuller 2012) and polarity detection (Pang and Lee 2008; Liu
2012; Wilson et al. 2005; Cambria 2016). While the former focuses on extracting a
set of emotion labels, the latter is usually a binary classification task with outputs
such as ‘positive’ versus ‘negative’, ‘thumbs up’ versus ‘thumbs down’ or ‘like’
versus ‘dislike’. These two tasks are highly inter-related and inter-dependent to the
extent that some sentiment categorization models, e.g., the Hourglass of Emotions
(Cambria et al. 2012), treat it as a unique task by inferring the polarity associated to
a sentence directly from the emotions this conveys. In many cases, in fact, emotion
recognition is considered a sub-task of polarity detection.
Polarity classification itself can also be viewed as a subtask of more advanced
analyses. For example, it can be applied to identifying ‘pro and con’ expressions that
can be used in individual reviews to evaluate the pros and cons that have influenced
the judgements of a product and that make such judgements more trustworthy.
Another instance of binary sentiment classification is agreement detection, that is,
given a pair of affective inputs, deciding whether they should receive the same or
differing sentiment-related labels.
Complementary to binary sentiment classification is the assignment of degrees of
positivity to the detected polarity or valence to the inferred emotions. If we waive the
assumption that the input under examination is opinionated and it is about one single
issue or item, new challenging tasks arise, e.g., subjectivity detection, opinion target
identification, and more (Cambria et al. 2015). The capability of distinguishing
whether an input is subjective or objective, in particular, can be highly beneficial
for a more effective sentiment classification. Moreover, a record can also have a
polarity without necessarily containing an opinion, for example a news article can
be classified into good or bad news without being subjective.
Typically, affective computing and sentiment analysis are performed over on-
topic documents, e.g., on the result of a topic-based search engine. However, several
studies suggested that managing these two task jointly can be beneficial for the
overall performances. For example, off-topic passages of a document could contain
irrelevant affective information and result misleading for the global sentiment
polarity about the main topic. Also, a document can contain material on multiple
topics that may be of interest to the user. In this case, it is therefore necessary to
identify the topics and separate the opinions associated with each of them.
Similar to topic detection is aspect extraction, a subtask of sentiment analysis
that consists in identifying opinion targets in opinionated text, i.e., in detecting
the specific aspects of a product or service the opinion holder is either praising
or complaining about. In a recent approach, Poria et al. (2016) used a 7-layer deep
convolutional neural network to tag each word in opinionated sentences as either
aspect or non-aspect word and developed a set of linguistic patterns for the same
purpose in combination with the neural network.
Other sentiment analysis subtasks include aspect extraction (Poria et al. 2016),
subjectivity detection (Chaturvedi et al. 2016), concept extraction (Rajagopal et al.
1 Affective Computing and Sentiment Analysis 5

2013), named entity recognition (Ma et al. 2016), and sarcasm detection (Poria et al.
2016), but also complementary tasks such as personality recognition (Poria et al.
2013), user profiling (Mihalcea and Garimella 2016) and especially multimodal
fusion (Poria et al. 2016). With increasing amounts of webcams installed in end-
user devices such as smart phones, touchpads, or netbooks, there is an increasing
amount of affective information posted to social online services in an audio or
audiovisual format rather than on a pure textual basis. For a rough impression on
the extent, consider that two days of video material are uploaded to YouTube on
average per minute. Besides speech-to-text recognition, this allows for additional
exploitation of acoustic information, facial expression and body movement analysis
or even the “mood” of the background music or the color filters, etc.
Multimodal fusion is to integrate all single modalities into a combined single
representation. There are basically two types of fusion techniques that have been
used in most of the literature to improve reliability in emotion recognition from
multimodal information: feature-level fusion and decision-level fusion (Konar and
Chakraborty 2015). The authors in Raaijmakers et al. (2008) fuse acoustic and
linguistic information. Yet, linguistic information is based on the transcript of the
spoken content rather than on automatic speech recognition output. In Morency et al.
(2011), acoustic, textual, and video features are combined for the assessment of
opinion polarity in 47 YouTube videos. A significant improvement is demonstrated
in a leave-one-video-out evaluation using Hidden-Markov-Models for classification.
As relevant features the authors identify polarized words, smile, gaze, pauses, and
voice pitch. Textual analysis is, however, also only based on the manual transcript
of spoken words.
In Poria et al. (2016), finally, the authors propose a novel methodology for
multimodal sentiment analysis, which consists in harvesting sentiments from Web
videos by demonstrating a model that uses audio, visual and textual modalities as
sources of information. They used both feature- and decision-level fusion methods
to merge affective information extracted from multiple modalities, achieving an
accuracy of nearly 80%.

1.3 General Categorization

Existing approaches to affective computing and sentiment analysis can be grouped

into three main categories: knowledge-based techniques, statistical methods, and
hybrid approaches.
Knowledge-based techniques are very popular because of their accessibility and
economy. Text is classified into affect categories based on the presence of fairly
unambiguous affect words like ‘happy’, ‘sad’, ‘afraid’, and ‘bored’. Popular sources
of affect words or multi-word expressions are Ortony’s Affective Lexicon (Ortony
et al. 1988), Wiebe’s linguistic annotation scheme (Wiebe et al. 2005), WordNet-
6 E. Cambria et al.

Affect (Strapparava and Valitutti 2004), SentiWordNet (Esuli and Sebastiani 2006),
SenticNet (Cambria et al. 2016), and other probabilistic knowledge bases trained
from linguistic corpora (Stevenson et al. 2007; Somasundaran et al. 2008; Rao
and Ravichandran 2009). The major weakness of knowledge-based approaches is
poor recognition of affect when linguistic rules are involved. For example, while
a knowledge base can correctly classify the sentence “today was a happy day”
as being happy, it is likely to fail on a sentence like “today wasn’t a happy
day at all”. To this end, more sophisticated knowledge-based approaches exploit
linguistics rules to distinguish how each specific knowledge base entry is used in
text (Poria et al. 2015). The validity of knowledge-based approaches, moreover,
heavily depends on the depth and breadth of the employed resources. Without
a comprehensive knowledge base that encompasses human knowledge, in fact,
it is not easy for a sentiment mining system to grasp the semantics associated
with natural language or human behavior. Another limitation of knowledge-based
approaches lies in the typicality of their knowledge representation, which is usually
strictly defined and does not allow handling different concept nuances, as the
inference of semantic and affective features associated with concepts is bounded
by the fixed, flat representation.
Statistical methods, such as support vector machines and deep learning, have
been popular for affect classification of texts and have been used by researchers
on projects such as Pang’s movie review classifier (Pang et al. 2002) and many
others (Hu and Liu 2004; Glorot et al. 2011; Socher et al. 2013; Lau et al. 2014;
Oneto et al. 2016). By feeding a machine learning algorithm a large training corpus
of affectively annotated texts, it is possible for the system to not only learn the
affective valence of affect keywords (as in the keyword spotting approach), but also
to take into account the valence of other arbitrary keywords (like lexical affinity)
and word co-occurrence frequencies. However, statistical methods are generally
semantically weak, i.e., lexical or co-occurrence elements in a statistical model have
little predictive value individually. As a result, statistical text classifiers only work
with acceptable accuracy when given a sufficiently large text input. So, while these
methods may be able to affectively classify user’s text on the page- or paragraph-
level, they do not work well on smaller text units such as sentences or clauses.
Hybrid approaches to affective computing and sentiment analysis, finally, exploit
both knowledge-based techniques and statistical methods to perform tasks such as
emotion recognition and polarity detection from text or multimodal data. Sentic
computing (Cambria and Hussain 2015), for example, exploits an ensemble of
knowledge-driven linguistic patterns and statistical methods to infer polarity from
text. Xia et al. (2015) used SenticNet and a Bayesian model for contextual
concept polarity disambiguation. Dragoni et al. (2014) proposed a fuzzy framework
which merges WordNet, ConceptNet and SenticNet to extract key concepts from a
sentence. iFeel (Araújo et al. 2014) is a system that allows users to create their own
sentiment analysis framework by combing SenticNet, SentiWordNet and other sen-
timent analysis methods. Chenlo and Losada (2014) used SenticNet to extract bag of
concepts and polarity features for subjectivity detection and other sentiment analysis
tasks. Chung et al. (2014) used SenticNet concepts as seeds and proposed a method
1 Affective Computing and Sentiment Analysis 7

of random walk in ConceptNet to retrieve more concepts along with polarity scores.
Other works propose the joint use of knowledge bases and machine learning for
Twitter sentiment analysis (Bravo-Marquez et al. 2014), short text message classifi-
cation (Gezici et al. 2013) and frame-based opinion mining (Recupero et al. 2014).

1.4 Conclusion

The passage from a read-only to a read-write Web made users more enthusiastic
about sharing their emotion and opinions through social networks, online com-
munities, blogs, wikis, and other online collaborative media. In recent years, this
collective intelligence has spread to many different areas of the Web, with particular
focus on fields related to our everyday life such as commerce, tourism, education,
and health.
Despite significant progress, however, affective computing and sentiment anal-
ysis are still finding their own voice as new inter-disciplinary fields. Engineers
and computer scientists use machine learning techniques for automatic affect
classification from video, voice, text, and physiology. Psychologists use their long
tradition of emotion research with their own discourse, models, and methods.
Affective computing and sentiment analysis are research fields inextricably bound
to the affective sciences that attempt to understand human emotions. Simply put, the
development of affect-sensitive systems cannot be divorced from the century-long
psychological research on emotion.
Hybrid approaches aim to better grasp the conceptual rules that govern sentiment
and the clues that can convey these concepts from realization to verbalization in
the human mind. In recent years, such approaches are gradually setting affective
computing and sentiment analysis as interdisciplinary fields in between mere
NLP and natural language understanding by gradually shifting from syntax-based
techniques to more and more semantics-aware frameworks Cambria and White
(2014), where both conceptual knowledge and sentence structure are taken into
account (Fig. 1.2).
So far, sentiment mining approaches from text or speech have been mainly based
on the bag-of-words model because, at first glance, the most basic unit of linguistic
structure appears to be the word. Single-word expressions, however, are just a subset
of concepts, multi-word expressions that carry specific semantics and sentics, that
is, the denotative and connotative information commonly associated with objects,
actions, events, and people. Sentics, in particular, specifies the affective information
associated with real-world entities, which is key for emotion recognition and
polarity detection, the basic tasks of affective computing and sentiment analysis.
The best way forward for these two fields, hence, is the ensemble application of
semantic knowledge and machine learning, where different approaches can cover
for each other’s flaws. In particular, the combined application of linguistics and
knowledge bases will allow sentiments to flow from concept to concept based on
8 E. Cambria et al.

Fig. 1.2 Jumping NLP curves

the dependency relation of the input sentence, while machine learning will act as
backup for missing concepts and unknown linguistic patterns.
Next-generation sentiment mining systems need broader and deeper common
and commonsense knowledge bases, together with more brain-inspired and
psychologically-motivated reasoning methods, in order to better understand
natural language opinions and, hence, more efficiently bridge the gap between
(unstructured) multimodal information and (structured) machine-processable data.
Looking ahead, blending scientific theories of emotion with the practical engi-
neering goals of analyzing sentiments in natural language and human behavior
will pave the way for development of more bio-inspired approaches to the design
of intelligent sentiment mining systems capable of handling semantic knowledge,
making analogies, learning new affective knowledge, and detecting, perceiving, and
‘feeling’ emotions.

References

Araújo, M., P. Gonçalves, M. Cha, and F. Benevenuto. 2014. iFeel: A system that compares and
combines sentiment analysis methods. In WWW, 75–78.
Bravo-Marquez, F., M. Mendoza, and B. Poblete. 2014. Meta-level sentiment models for big social
data analysis. Knowledge-Based Systems 69: 86–99.
1 Affective Computing and Sentiment Analysis 9

Calvo, R., and S. D’Mello. 2010. Affect detection: An interdisciplinary review of models, methods,
and their applications. IEEE Transactions on Affective Computing 1(1): 18–37.
Cambria, E. 2016. Affective computing and sentiment analysis. IEEE Intelligent Systems 31(2):
102–107.
Cambria, E., and A. Hussain. 2015. Sentic computing: A common-sense-based framework for
concept-level sentiment analysis. Cham: Springer.
Cambria, E., A. Livingstone, and A. Hussain. 2012. The hourglass of emotions. In Cognitive
behavioral systems, ed. A. Esposito, A. Vinciarelli, and R. Hoffmann, V. Muller, Lecture notes
in computer science, vol. 7403, 144–157. Berlin/Heidelberg: Springer.
Cambria, E., S. Poria, R. Bajpai, and B. Schuller. 2016. SenticNet 4: A semantic resource for
sentiment analysis based on conceptual primitives. In COLING, 2666–2677.
Cambria, E., S. Poria, F. Bisio, R. Bajpai, and I. Chaturvedi. 2015. The CLSA model: A novel
framework for concept-level sentiment analysis. In Computational linguistics and intelligent
text processing. CICLing 2015, ed. A. Gelbukh, LNCS, vol. 9042, 3–22. Cham: Springer.
Cambria, E., H. Wang, and B. White. 2014. Guest editorial: Big social data analysis. Knowledge-
Based Systems 69: 1–2.
Cambria, E., and B. White. 2014. Jumping NLP curves: A review of natural language processing
research. IEEE Computational Intelligence Magazine 9(2): 48–57.
Chaturvedi, I., E. Cambria, and D. Vilares. 2016. Lyapunov filtering of objectivity for Spanish
sentiment model. In: IJCNN, 4474–4481.
Chenlo, J.M., and D.E. Losada. 2014. An empirical study of sentence features for subjectivity and
polarity classification. Information Sciences 280: 275–288.
Chung, J.K.C., C.E. Wu, and R.T.H. Tsai. 2014. Improve polarity detection of online reviews with
bag-of-sentimental-concepts. In Proceedings of the 11th ESWC. Semantic Web Evaluation
Challenge. Crete: Springer.
Dragoni, M., A.G. Tettamanzi, and C. da Costa Pereira. 2014. A fuzzy system for concept-level
sentiment analysis. In Semantic web evaluation challenge, 21–27. Cham: Springer.
Esuli, A., and F. Sebastiani. 2006. SentiWordNet: A publicly available lexical resource for opinion
mining. In LREC.
Gezici, G., R. Dehkharghani, B. Yanikoglu, D. Tapucu, and Y. Saygin. 2013. Su-sentilab: A
classification system for sentiment analysis in twitter. In International Workshop on Semantic
Evaluation, 471–477.
Glorot, X., A. Bordes, and Y. Bengio. 2011. Domain adaptation for large-scale sentiment
classification: A deep learning approach. In ICML, Bellevue.
Gunes, H., and B. Schuller. 2012. Categorical and dimensional affect analysis in continuous input:
Current trends and future directions. Image and Vision Computing 31(2): 120–136.
Hu, M., and B. Liu. 2004. Mining and summarizing customer reviews. In KDD, Seattle.
Konar, A., and A. Chakraborty. 2015. Emotion recognition: A pattern analysis approach. Hoboken:
Wiley & Sons.
Lau, R., Y. Xia, and Y. Ye. 2014. A probabilistic generative model for mining cybercriminal
networks from online social media. IEEE Computational Intelligence Magazine 9(1): 31–43
Liu, B. 2012. Sentiment analysis and opinion mining. San Rafael: Morgan and Claypool.
Ma, Y., E. Cambria, and S. Gao. 2016. Label embedding for zero-shot fine-grained named entity
typing. In COLING, Osaka, 171–180.
Mihalcea, R., and A. Garimella. 2016. What men say, what women hear: Finding gender-specific
meaning shades. IEEE Intelligent Systems 31(4): 62–67.
Minsky, M. 2006. The emotion machine: Commonsense thinking, artificial intelligence, and the
future of the human mind. New York: Simon & Schuster.
Morency, L.P., R. Mihalcea, and P. Doshi. 2011. Towards multimodal sentiment analysis:
Harvesting opinions from the web. In International Conference on Multimodal Interfaces
(ICMI), 169–176. New York: ACM.
Oneto, L., F. Bisio, E. Cambria, and D. Anguita. 2016. Statistical learning theory and ELM for big
social data analysis. IEEE Computational Intelligence Magazine 11(3): 45–55.
10 E. Cambria et al.

Ortony, A., G. Clore, and A. Collins. 1988. The cognitive structure of emotions. Cambridge:
Cambridge University Press.
Pang, B., and L. Lee. 2008. Opinion mining and sentiment analysis. Foundations and Trends in
Information Retrieval 2: 1–135.
Pang, B., L. Lee, and S. Vaithyanathan. 2002. Thumbs up? Sentiment classification using machine
learning techniques. In EMNLP, Philadelphia, 79–86.
Pantic, M., N. Sebe, J. Cohn, and T. Huang. 2005. Affective multimodal human-computer
interaction. In ACM International Conference on Multimedia, New York, 669–676.
Picard, R. 1997. Affective computing. Boston: The MIT Press.
Poria, S., E. Cambria, and A. Gelbukh. 2016. Aspect extraction for opinion mining with a deep
convolutional neural network. Knowledge-Based Systems 108: 42–49.
Poria, S., E. Cambria, A. Gelbukh, F. Bisio, and A. Hussain. 2015. Sentiment data flow analysis
by means of dynamic linguistic patterns. IEEE Computational Intelligence Magazine 10(4):
26–36.
Poria, S., E. Cambria, D. Hazarika, and P. Vij. 2016. A deeper look into sarcastic tweets using deep
convolutional neural networks. In COLING, 1601–1612.
Poria, S., E. Cambria, N. Howard, G.B. Huang, and A. Hussain. 2016. Fusing audio, visual and
textual clues for sentiment analysis from multimodal content. Neurocomputing 174: 50–59.
Poria, S., I. Chaturvedi, E. Cambria, and A. Hussain. 2016. Convolutional MKL based multimodal
emotion recognition and sentiment analysis. In ICDM, 439–448.
Poria, S., A. Gelbukh, B. Agarwal, E. Cambria, and N. Howard. 2013. Common sense knowledge
based personality recognition from text. In Advances in soft computing and its applications,
484–496. Berlin/Heidelberg: Springer.
Raaijmakers, S., K. Truong, and T. Wilson. 2008. Multimodal subjectivity analysis of multiparty
conversation. In EMNLP, Edinburgh, 466–474.
Rajagopal, D., E. Cambria, D. Olsher, and K. Kwok. 2013. A graph-based approach to common-
sense concept extraction and semantic similarity detection. In WWW, Rio De Janeiro, 565–570.
Rao, D., and D. Ravichandran. 2009. Semi-supervised polarity lexicon induction. In EACL,
Athens, 675–682.
Recupero, D.R., V. Presutti, S. Consoli, A. Gangemi, and A. Nuzzolese. 2014. Sentilo: Frame-
based sentiment analysis. Cognitive Computation 7(2): 211–225.
Schuller, B., A. Batliner, S. Steidl, and D. Seppi. 2011. Recognising realistic emotions and affect
in speech: State of the art and lessons learnt from the first challenge. Speech Communication
53(9/10): 1062–1087.
Socher, R., A. Perelygin, J.Y. Wu, J. Chuang, C.D. Manning, A.Y. Ng, and C. Potts. 2013.
Recursive deep models for semantic compositionality over a sentiment treebank. In EMNLP,
1642–1654.
Somasundaran, S., J. Wiebe, and J. Ruppenhofer. 2008. Discourse level opinion interpretation. In
COLING, Manchester, 801–808.
Stevenson, R., J. Mikels, and T. James. 2007. Characterization of the affective norms for english
words by discrete emotional categories. Behavior Research Methods 39: 1020–1024.
Strapparava, C., and A. Valitutti. 2004. WordNet-Affect: An affective extension of WordNet. In
LREC, Lisbon, 1083–1086.
Wiebe, J., T. Wilson, and C. Cardie. 2005. Annotating expressions of opinions and emotions in
language. Language Resources and Evaluation 39(2): 165–210.
Wilson, T., J. Wiebe, and P. Hoffmann. 2005. Recognizing contextual polarity in phrase-level
sentiment analysis. In HLT/EMNLP, Vancouver, 347–354.
Xia, Y., E. Cambria, A. Hussain, and H. Zhao. 2015. Word polarity disambiguation using bayesian
model and opinion-level features. Cognitive Computation 7(3): 369–380.
Zeng, Z., M. Pantic, G. Roisman, and T. Huang. 2009. A survey of affect recognition methods:
Audio, visual, and spontaneous expressions. IEEE Transactions on Pattern Analysis and
Machine Intelligence 31(1): 39–58.