Singh et al. Hum. Cent. Comput. Inf. Sci.

(2017) 7:32
DOI 10.1186/s13673-017-0116-3

RESEARCH Open Access

Optimization of sentiment analysis using

machine learning classifiers
Jaspreet Singh* , Gurvinder Singh and Rajinder Singh

*Correspondence:
profjaspreetbatth@gmail. Abstract
com Words and phrases bespeak the perspectives of people about products, services,
Department of Computer
Science, Guru Nanak Dev governments and events on social media. Extricating positive or negative polarities
University, Amritsar, India from social media text denominates task of sentiment analysis in the field of natural
language processing. The exponential growth of demands for business organiza-
tions and governments, impel researchers to accomplish their research in sentiment
analysis. This paper leverages four state-of-the-art machine learning classifiers viz. Naïve
Bayes, J48, BFTree and OneR for optimization of sentiment analysis. The experiments
are performed using three manually compiled datasets; two of them are captured
from Amazon and one dataset is assembled from IMDB movie reviews. The efficacies
of these four classification techniques are examined and compared. The Naïve Bayes
found to be quite fast in learning whereas OneR seems more promising in generating
the accuracy of 91.3% in precision, 97% in F-measure and 92.34% in correctly classified
instances.
Keywords: Sentiment analysis, Social media text, Movie reviews, Product reviews,
Machine learning classifiers

Introduction to sentiment analysis

The popularity of rapidly growing online social networks and electronic media based
societies has influenced the young researchers to pursue their work on sentiment analy-
sis. These days organizations quite keen assess their customers or public opinion about
their products from social media text [1]. The online service providers are hooked on
assessing social media data on blogs, online forums, comments, tweets and product
reviews. This assessment is exploited for their decision making or amelioration of their
services or quality of products. The applications of sentiment analysis encompass the
areas like social event planning, election campaigning, healthcare monitoring, consumer
products and awareness services. The immoderate use of internet by business organiza-
tions all around the globe has noticed that opinionated web text has molded our business
plays and socio-economic systems. The computational power is fueled by burgeon of
machine learning techniques. This work focused on four text classifiers utilized for senti-
ment analysis viz. Naïve Bayes, J48, BFTree and OneR algorithm. The “Machine learning
techniques for sentiment analysis” section of this paper provides the intuition behind the
task of sentiment classification by leveraging the modeling of aforementioned four clas-
sifiers. The architecture of proposed model using four sentiment classifiers is disposed

© The Author(s) 2017. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License
(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and
indicate if changes were made.
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 2 of 12

in “Proposed methodology for optimization of sentiment prediction using weka” sec-

tion. The related work with recent contributions of machine learning in the field of senti-
ment classification is described in “Related work” section. In “Datasets taken” section,
the three manually annotated datasets are described along with their preprocessing. The
experimental results and discussion of efficacies of classifiers are cataloged in “Results
and discussions” section followed by the ending remarks along with a future direction in
“Conclusion” section.

Levels of sentiment
Due to scarcity of opinion text available in digital form, very less research interest on
computational linguistics in the last decade of twentieth century was witnessed [2–4].
The escalation of social media text on internet attracts young researchers to define the
level of granularities of text. The web text is classified into three levels viz. document
level, sentence level and word level. In [5], the fourth level granularity is defined by using
deep convolution neural network. This fourth level is character level feature extrac-
tion approach used for extracting features of each character window from given word
(Table 1).

Machine learning techniques for sentiment analysis

The social networking sites dispense their data conveniently and freely on the web. This
availability of data entices the interest of young researchers to plunge them in the field of
sentiment analysis. People express their emotions and perspectives on the social media
discussion forums [6]. The business organizations employ researchers to investigate the
unrevealed facts about their products and services. Spontaneous and automatic deter-
mination of sentiments from reviews is the main concern of multinational organizations
[7–10]. The machine learning techniques have improved accuracy of sentiment analysis
and expedite automatic evaluation of data these days. This work attempted to utilize four
machine learning techniques for the task of sentiment analysis. The modeling of four
techniques is briefly discussed below.

Naïve Bayes used for sentiment classification

The dichotomy of sentiment is generally decided by the mindset of an author of text
whether he is positively or negatively oriented towards his saying [6, 11–13]. Naïve Bayes

Table 1 Levels of sentiment along with their attributes

Level Delimiter Depth of Granu- Multiplicity of sen- Interpretation
larity timents of sentiments

1. Document [15] ‘\n’ Newline char- Overall opinion at Single opinion of Overall sentiment of
acter upper level multiple entities one document
2. Sentence [16] ‘.’ Period character Factual polarity of Multiple opinions of Subjectivity clas-
individual sen- multiple entities sification
tences
3. Entity or aspect Space character or At finest level words Single opinion of Two-tuple as <Senti-
level [12] named entities are the target single entity ment, target>
entities
4. Character level [5] Special symbols and Micro level of char- Multiple opinions Morphological
space characters acter embedding about single word extraction of words
are omitted entity
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 3 of 12

classifier is a popular supervised classifier, furnishes a way to express positive, negative

and neutral feelings in the web text. Naïve Bayes classifier utilizes conditional probabil-
ity to classify words into their respective categories. The benefit of using Naïve Bayes
on text classification is that it needs small dataset for training. The raw data from web
undergoes preprocessing, removal of numeric, foreign words, html tags and special
symbols yielding the set of words. The tagging of words with labels of positive, negative
and neutral tags is manually performed by human experts. This preprocessing produces
word-category pairs for training set. Consider a word ‘y’ from test set (unlabeled word
set) and a window of n-words ­(x1, ­x2, …… ­xn) from a document. The conditional proba-
bility of given data point ‘y’ to be in the category of n-words from training set is given by:

n
   P(xi /y)
P(y/x1 , x2 , . . . . . . xn ) = P y × (1)
P(x1 , x2 , . . . . . . xn )
i=1

Consider an example of a movie review for movie “Exposed”. The experimentation

with Naïve Bayes yields the following results.

J48 algorithm used for sentiment prediction

The hierarchical mechanism divides feature space into distinct regions followed by the
categorization of sample into category labels. J48 is a decision tree based classifier used
to generate rules for the prediction of target terms. It has an ability to deal with larger
training datasets than other classifiers [14]. The word features for sentences of corpus
taken from labeled arff file of training set are represented in the leaf nodes of decision
tree. In the test set every time when a near feature qualifies the label condition of inter-
nal feature node, its level is lifted up in the same branch of decision tree. The assignment
of labels to the word features of test set gradually generates different two branches of
decision tree. J48 algorithm uses entropy function for testing the classification of terms
from the test set.
n
 |Termj| |Termj|
Entropy(Term) = − log2 (2)
|Term| |Term|
j=1

where (Term) can be unigram, bigram and trigram. In this study we have considered
unigrams and bigrams. The example in the Table 2 contains bigrams like “Horrible act-
ing”, “Bad writing” and “Very misleading” are labeled with negative sentiment whereas
the term “More enjoyable” reflects positive sentiment towards the movie. The decision
tree of J48 algorithm for obtaining sentiment form text is represented in the Fig. 1 below.

BFTREE algorithm used for sentiment prediction

Another classification approach outperforms J48, C4.5 and CART by expanding only
best node in the depth first order. BFTree algorithm excavates the training file for locat-
ing best supporting matches of positive and negative terms in the test file. BFTree algo-
rithm keeps heuristic information gain to identify best node by probing all collected
word features. The only difference between J48 and BFTree classifier is the computation
order in which decision tree is built. The decision tree disparate feature terms of plain
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 4 of 12

Table 2 Initial four reviews of training set and two reviews test set
Set Sentence Review Class

Train 1 Horrible acting and bad writing Neg

2 I never did work out what the dog death scene was all about! Neg
3 The trailer is exciting but very misleading Neg
4 The basic structure of storyline is good Pos
Test 1 The effective start and a very detective story Pos
2 An evening locked up in the toilet is more enjoyable Pos
Second review of test set is negative but Naïve Bayes is lacking in context based sentiment classification

Fig. 1 J48’s Decision Tree for terms of Example in Table 2

text taken from movie reviews and classify them at document level by tagging appro-
priate labels. BFTree extracts best node from labeled and trained binary tree nodes to
reduce the error computed from information gain.

 |Si |
Infogain (S, A) = Entropy(S) − × Entropy(Si ) (3)
S
i∈V (A)

where S is word feature term of test set and A is the attribute of sampled term from
training set. V(A) denotes set of all possible values of A. The binary tree stops growing
when an attribute A captures single value or when value of information gain vanishes.

OneR algorithm used for sentiment prediction

OneR algorithm is a classification approach which restricts decision tree to level one
thereby generating one rule. One rule makes prediction on word feature terms with min-
imal error rate due to repetitive assessment of word occurrences. The classification of
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 5 of 12

most frequent terms of a particular sentence is made on the basis of class of featured
terms from training set. The demonstration of OneR algorithm for sentiment prediction
with smallest error of classification is given below:

Step 1 Select a featured term from training set.

Step 2 Train a model using step 3 and step 4.
Step 3 For each prediction term.
For each value of that predictor.
Count frequency of each value of target term.
Find most frequent class.
Make a rule and assign that class to predictor.
Step 4 Calculate total error of rules of each predictor.
Step 5 Choose predictor with smallest error.

Proposed methodology for optimization of sentiment prediction using weka

The preprocessing of raw text from web is done in python 3.5 using NLTK and bs4
libraries. Each review in the first dataset is parsed with NLTK’s parser and title of the
review is considered as a feature. We have obtained 15 features from first dataset and 42
features from each of second and third dataset. The CSV files generated from Python are
converted to ARFF files for WEKA 3.8. Only two sentiment labels namely Pos for posi-
tive and Neg for negative are used for assigning sentences. The working methodology of
proposed work for optimization of sentiment prediction is given below in Fig. 2.
After loading files with ARFF loader, the class assigner picks up appropriate class
labels from dataset and performs feature selection on the basis of frequently used head-
ings and most frequent titles. The feature selector module is implemented using three
feature selection methods namely Document Frequency (DF), Mutual Information (MI)
and Information Gain (IG). The mathematical modeling of these feature selection meth-
ods requires some probability distributions and statistical notations described below:

Fig. 2 Proposed methodology

Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 6 of 12

P(w): Probability that a document ‘d’ contains term ‘w’.

P(c’): Probability that document ‘d’ does not belongs to category ‘c’.
P(w, c): Joint probability that document ‘d’ contains word term ‘w’ of category ‘c’.
P(c/w): Conditional probability that a document ‘d’ belongs to category ‘c’ under the
condition that ‘d’ contains word term ‘w’.
Similarly other notations like P(w’), P(w/c), P(w/c’), P(c/w’) and P(c’/w) are taken and
{c} is the set of categories.
N1: Number of documents that exhibit category ‘c’ and contain term ‘w’.
N2: Number of documents that do not belong to category ‘c’ but contains term ‘w’.
N3: Number of documents that belong to category ‘c’ and do not contain term ‘w’.
N4: Number of documents that neither belong to category ‘c’ nor contain term ‘w’.
N: Total number of document reviews.
DF method qualifies only those documents in which a higher frequency terms are
considered.

m

DF = N1i (4)
i=1

The MI method measures features of text by computing similarity of word terms ‘w’
and category ‘c’.

P(w/c)
SimInfo (w, c) = log (5)
P(w)

N1 × N
MI = log (6)
(N1 + N3 )(N1 + N2 )

The IG-construct measures similarity information for category by exploiting probabili-

ties of absence or presence of terms in a document review.
  
IG(w) = − P(c) · log P(c) + P(w) P(c/w) · log P(c/w)
 ′   ′
  ′
 (7)
+P w P c/w · log P c/w

The normalization module converts all letters into lowercase, removal of punctuation
marks and special symbols, conversion of numbers into words, expansion of abbrevi-
ation and limiting the average length of twenty words in a sentence. Each sentence is
delimited by a newline character. The Python’s NLTK and bs4 libraries are used for this
purpose. Data splitter take the ratio of (80:20) of (Train: Test) subsets. We have used
manual splitting of dataset at the time of retrieval of data from web. The four classifi-
ers are trained with training subsets followed by performance evaluation. The evaluation
metrics taken in the experiment are precision, recall, accuracy and F-measure.
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 7 of 12

Related work
Existing approaches of sentiment prediction and optimization widely includes SVM and
Naïve Bayes classifiers. Hierarchical machine learning approaches yields moderate per-
formance in classification tasks whereas SVM and Multinomial Naïve Bayes are proved
better in terms of accuracy and optimization. Sentiment analysis using neural network
architectures has appeared in very few works. The sentiment prediction methods using
recursive neural networks and deep convolution neural networks are bit complex in cap-
turing compositionality of words. Extracting character level features and embeddings
of complex words is found hard in many neural network architectures whereas extract-
ing sentence level or word level features such as morphological tags and stems are more
effectively achieved in convolutional neural networks. A very few researchers have used
J48, BFTree and OneR for the task of sentiment prediction. These three classifiers are
utilized for other classification tasks like emotion recognition from text and twitter’s text
categorizations. The summary of benchmarks related to machine learning techniques in
terms of accuracy of classification is listed in the Table 2. SVM and Naive Bayes are prov-
ing better in terms of benchmarks than other machine learning techniques (Table 3).

Datasets taken
Three Datasets are manually annotated from http://www.amazon.in. First dataset con-
sists of product reviews of Woodland’s wallet are taken from 12th October 2016 to 25th
October 2016 for training set containing 88 reviews and from 25th October 2016 to 30th
October 2016 for testing set containing 12 randomly chosen product reviews with their
sentiments prediction using four machine learning algorithms. Second dataset consists
of 7465 Digital Camera reviews of Sony are taken from 01st October 2016 to 25th Octo-
ber 2016 for training set and 1000 reviews are from 25th October 2016 to 30th October
2016 for test dataset. Third dataset consists of movie reviews taken from http://www.
imdb.com. It contains 2421 reviews for training set and 500 reviews for test set.

Results and discussions

The experiment is carried out by using freeware WEKA software tool for classification
of sentiments in the text. Standard implementations of Naïve Bayes, J48, BFTree and
OneR algorithms are exploited from WEKA version 3.8. The classification accuracy of
first dataset shows 100% classification accuracy with Naïve Bayes in some of the epochs
because of small size of dataset. The average of 29 epochs for all four classifiers on sec-
ond and third datasets is presented in Table 4 below. Naïve Bayes shows faster learning
among four classifiers whereas J48 found to be slower. OneR classifier is leading from
other three classifiers in percentage of correctly classified instances. The accuracy of J48
algorithm is promising in true positive and false positive rates.
Results of classification accuracies for the test subsets with 42 and 15 attributes are
recorded. The average accuracies of 29 runs on three datasets is presented in Table 5
Table 3 Benchmarks of classifier’s accuracies
Author(/s) (year of publication) Classifiers and features used Description Accuracy of classification (%age)

Kiritchenko and Mohammad (2016) SVM with RBF kernel, POS, sentiment score, emoti- Supervised sentiment analysis system using 82.60 for bigrams 80.90 for trigrams
cons, embedding vectors [17] real-valued sentiment score to analyze social
networking data
Dashtipour et al. (2016) SVM, MNB, maximum entropy [18] Multilingual sentiment analysis for improving the 86.35
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32

performance of sentiment classification

Tan and Zhang (2008) Naive Bayes, SVM and k-NN [15] A text classification system for sentiment detec- 82
tion of Chinese documents
Mohammad et al. (2015) Automatic SVM elongated words, emoticons, Automatic emotion detection system for 2012 US 56.84
negation feature, position feature. [16] presidential election tweets
Sobhani et al. (2016) Linear kernel SVM, n-grams and word embeddings Stance and sentiment detection system 70.3
[19]
Poria et al. (2014) Maximum entropy, naive Bayes and SVM, ELM [20] Concept level sentiment analysis for movie review 67.35 for ELM and 65.67 for SVM
dataset
Socher (2016) SVM, Naive Bayes [21] Deep learning for sentiment analysis 85.4
Turney and Mohammad (2014) Lexicon based entailment, SVM [22] Proposed three algorithms namely balAPinc, 68.70 for balAPinc 70.20 for ConVecs 74.50 for
ConVecs and SimDiffs. Tested on three different SimDiffs
datasets i.e. KDSZ dataset created by Kotler-
man et al. (2010), BBDS dataset introduced by
Baroni et al. (2012) and JMTH dataset created by
Jurgens et al. (2012) in SemEval-2012 Task2
Mohammad et al. (SemEval-2016) task 6 SVM, unigrams, n-grams, hashtags, combined Automatic stance detection system from tweets. Favg = 67.82
features [23] Where team MITRE has achieved highest level
in accuracy
Cernian et al. (2015) POS, SentiWordNet, Sentiment Score and Synset Proposed framework for sentiment analysis is 61
[7] tested on 300 product reviews from Amazon
Page 8 of 12
Table 3 continued
Author(/s) (year of publication) Classifiers and features used Description Accuracy of classification (%age)
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32

Kiritchenko et al. (SemEval-2016) task 7 Supervised learning, random forest, PMI, Gaussian Automatic sentiment score determination model Kendall’s rank coeff. (K) K = 0.704 for Gen. English
regression, NRC emoticons, SentiWordNet [6] for general English, English Twitter corpus and 0.523 for English Twitter, and 0.536 for Arabic
Arabic Twitter corpus Twitter
Pang et al. (2002) Naïve Bayes, SVM, maximum entropy classifiers Performed feature based analysis on movie 78.7 for NB, 77.7 for ME and 82.9 for SVM
with unigrams, bigrams, POS, adjectives and reviews using three machine learning classifiers
word frequency features [3] for sentiment classification
Nogueira dos Santos and Gatti (2014) Convolution neural network using word-level and Proposed convolution neural network for classifi- 85.7 for binary classification, 48.3 for fine grained
character-level embedding vectors [5] cation of short text messages from Twitter using classification and 86.4 for STS corpus
character level word embeddings
Poria et al. (2015) Ensemble classifier using POS, Sentic, negation, The proposed algorithm captures contextual 88.12 for movie review dataset, 88.27 for Blitzer
modification and common sense knowledge polarity and flow of concepts from text for derived dataset and 82.75 for Amazon corpus
features [4] dynamic sentiment analysis
Poria et al. (2016) SVM and Naïve Bayes, CNN used for extracting Convolutional multiple kernel learning for enhanc- 96.12 for proposed model without feature selection
video, audio and textual features (word embed- ing the performance of sentiment analysis and and 96.55 with feature selection
dings and POS) [24] emotion recognition
Wang et al. (2016) Back propagation and stochastic gradient descent Proposed regional convolutional neural network Pearson correlation coefficient r = 0.778 between
used to learn model parameters along with and long short term memory model for fine CNN-LSTM and LSTM, for English text and
features such as n-gram and word vector for grained sentiment analysis r = 0.781 for Chinese text
Valence–Arousal prediction [25]
Page 9 of 12
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 10 of 12

below. All four classifiers improved in accuracies with the increase of features from 15 to
42. This shows the direct proportionality of multiple features with learning capability for
machine learning algorithms.

Conclusion
This paper exploits four machine learning classifiers for sentiment analysis using three
manually annotated datasets. The mean of 29 epochs of experimentation recorded in
Table 4 shows that OneR is more precise in terms of percentage of correctly classified
instances. On the other hand, Naïve Bayes exhibits faster learning rate and J48 reveals
adequacy in the true positive and false positive rates. Table 5 reveals the truth that J48
and OneR are better for smaller dataset of woodland’s wallet reviews. The preprocessing
of proposed methodology is limited to extract foreign words, emoticons and elongated
words with their appropriate sentiments. The future work in the task of sentiment analy-
sis has scope to improve preprocessing with word embeddings using deep neural net-
works and can also extend this study through convolution neural networks.

Table 4 Performance evaluation of four classifiers

Classifiers Time taken Correctly Incorrectly Accuracy Accuracy Precision F-measure
(s) classified classified TP rate FP rate
instances instances

Naïve Bayes 7.79 85.24 14.61 0.456 0.134 0.831 0.812

J-48 49.73 89.73 11.33 0.967 0.003 0.877 0.917
BFTree 21.12 90.07 9.03 0.892 0.025 0.883 0.721
OneR 24.45 92.34 8.66 0.9 0.061 0.913 0.97

Table 5 Test accuracies of classification algorithms for three datasets

Classification algorithm Data set Test accuracy with 42 Test accuracy with 15
features of text (%age) features of text (%age)

Naïve Bayes D1 85.127 78.814

D2 73.517 69.298
D3 76.782 68.864
J-48 D1 87.622 85.452
D2 76.865 63.454
D3 74.231 69.895
BFTree D1 84.982 80.232
D2 69.452 65.156
D3 61.512 60.564
OneR D1 87.652 84.452
D2 76.563 72.788
D3 65.876 63.521
D1: Woodland’s wallet reviews, D2: Sony digital camera reviews, D3: IMDB movie reviews
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 11 of 12

Authors’ contributions
JS made substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data.
GS helped in revision and has given final approval of the version to be published. RS agreed to be accountable for all
aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropri-
ately investigated and resolved. All authors read and approved the final manuscript.

Acknowledgements
This research was supported by Department of Computer Science, Guru Nanak Dev University, Amritsar. I thank Dr.
Gurvinder Singh and Dr. Rajinder Singh for their participation in experimental work and their assistance to improve the
manuscript.

Competing interests
This research work has non-financial Academic and intellectual competing interests.

Availability of data and materials

I submit that I can make the experimental data and materials available after the completion of my thesis.

Consent for publication

We hereby grant and assign all rights to Human-centric Computing and Information Sciences for publication.

Ethics approval and consent to participate

The article submitted is an original work and has neither been published in any other peer-reviewed journal nor under
consideration for publication by any other journal.

Funding information
No funding was received from any funder.

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Received: 24 December 2016 Accepted: 4 September 2017

References
1. Parvathy G, Bindhu JS (2016) A probabilistic generative model for mining cybercriminal network from online social
media: a review. Int J Comput Appl 134(14):1–4. doi:10.5120/ijca2016908121
2. Cambria E, White B (2014) Jumping NLP curves: a review of natural language processing research. IEEE Comput
Intell Mag 9(2):48–57. doi:10.1109/mci.2014.2307227
3. Pang B, Lee L, Vaithyanathan S (2002) Thumbs up? In: Proceedings of the ACL-02 conference on empirical methods
in natural language processing—EMNLP ‘02. doi:10.3115/1118693.1118704
4. Poria S, Cambria E, Gelbukh A, Bisio F, Hussain A (2015) Sentiment data flow analysis by means of dynamic linguistic
patterns. IEEE Comput Intell Mag 10(4):26–36. doi:10.1109/mci.2015.2471215
5. Nogueira dos Santos C, Gatti M (2014) Deep convolution neural networks for sentiment analysis of short texts. In:
Proceedings of COLING 2014, the 25th international conference on computational linguistics. p 69–78
6. Kiritchenko S, Mohammad S, Salameh M (2016) SemEval-2016 task 7: determining sentiment intensity of English
and Arabic phrases. In: Proceedings of the 10th international workshop on semantic evaluation (SemEval-2016).
doi:10.18653/v1/s16-1004
7. Cernian A, Sgarciu V, Martin B (2015) Sentiment analysis from product reviews using SentiWordNet as lexi-
cal resource. In: 2015 7th international conference on electronics, computers and artificial intelligence (ECAI).
doi:10.1109/ecai.2015.7301224
8. Hammer HL, Solberg PE, Øvrelid L (2014) Sentiment classification of online political discussions: a comparison of a
word-based and dependency-based method. In: Proceedings of the 5th workshop on computational approaches
to subjectivity, sentiment and social media analysis. doi:10.3115/v1/w14-2616
9. Zadeh L (2006) Toward human-level machine intelligence. In: 2006 18th IEEE international conference on tools with
artificial intelligence (ICTAI’06). doi:10.1109/ictai.2006.114
10. Joachims T (2002) Text classification. Learning to classify text using support vector machines. p 7-33.
doi:10.1007/978-1-4615-0907-3_2
11. Wanton TM, Porrata AP, Guijarro AM, Balahur A (2010) Opinion polarity detection—using word sense disambigua-
tion to determine the polarity of opinions. In: Proceedings of the 2nd international conference on agents and
artificial intelligence. doi:10.5220/0002703504830486
12. Xia Y, Cambria E, Hussain A, Zhao H (2014) Word polarity disambiguation using bayesian model and opinion-level
features. Cogn Comput 7(3):369–380. doi:10.1007/s12559-014-9298-4
13. Dey L, Chakraborty S, Biswas A, Bose B, Tiwari S (2016) Sentiment analysis of review datasets using Naïve Bayes’ and
K-NN classifier. Int J Inform Eng Electron Bus 8(4):54–62. doi:10.5815/ijieeb.2016.04.07
14. Nie CY, Wang J, He F, Sato R (2015) Application of J48 decision tree classifier in emotion recognition based on chaos
characteristics. In: Proceedings of the 2015 international conference on automation, mechanical control and com-
putational engineering. doi:10.2991/amcce-15.2015.330
15. Tan S, Zhang J (2008) An empirical study of sentiment analysis for Chinese documents. Expert Syst Appl 34(4):2622–
2629. doi:10.1016/j.eswa.2007.05.028
Singh et al. Hum. Cent. Comput. Inf. Sci. (2017) 7:32 Page 12 of 12

16. Mohammad SM, Zhu X, Kiritchenko S, Martin J (2015) Sentiment, emotion, purpose, and style in electoral tweets. Inf
Process Manage 51(4):480–499. doi:10.1016/j.ipm.2014.09.003
17. Kiritchenko S, Mohammad SM (2016) sentiment composition of words with opposing polarities. In: Proceedings
of the 2016 conference of the north american chapter of the association for computational linguistics: human
language technologies. doi:10.18653/v1/n16-1128
18. Dashtipour K, Poria S, Hussain A, Cambria E, Hawalah AY, Gelbukh A, Zhou Q (2016) Multilingual sentiment
analysis: state of the art and independent comparison of techniques. Cogn Comput 8(4):757–771. doi:10.1007/
s12559-016-9415-7
19. Sobhani P, Mohammad S, Kiritchenko S (2016) Detecting stance in tweets and analyzing its interaction with
sentiment. In: Proceedings of the 5th joint conference on lexical and computational semantics. doi:10.18653/v1/
s16-2021
20. Poria S, Cambria E, Winterstein G, Huang G (2014) Sentic patterns: dependency-based rules for concept-level senti-
ment analysis. Knowl Based Syst 69:45–63. doi:10.1016/j.knosys.2014.05.005
21. Socher R (2016) deep learning for sentiment analysis—invited talk. In: Proceedings of the 7th workshop on compu-
tational approaches to subjectivity, sentiment and social media analysis. doi:10.18653/v1/w16-0408
22. Turney PD, Mohammad SM (2014) Experiments with three approaches to recognizing lexical entailment. Nat Lang
Eng 21(03):437–476. doi:10.1017/s1351324913000387
23. Mohammad S, Kiritchenko S, Sobhani P, Zhu X, Cherry C (2016) SemEval-2016 task 6: detecting stance in tweets. In:
Proceedings of the 10th international workshop on semantic evaluation (SemEval-2016). doi:10.18653/v1/s16-1003
24. Poria S, Chaturvedi I, Cambria E, Hussain A (2016) Convolutional MKL based multimodal emotion recognition and
sentiment analysis. In: 2016 IEEE 16th international conference on data mining (ICDM). doi:10.1109/icdm.2016.0055
25. Wang J, Yu L, Lai KR, Zhang X (2016) Dimensional sentiment analysis using a regional CNN-LSTM model. In: Proceed-
ings of the 54th annual meeting of the association for computational linguistics, vol 2: short papers. doi:10.18653/
v1/p16-2037