SARCASM AS A CONTRADICTION BETWEEN A TWEET AND ITS TEMPORAL FACTS: A PATTERNBASED APPROACH

International Journal on Natural Language Computing (IJNLC) Vol.7, No.5, October 2018
DOI: 10.5121/ijnlc.2018.7507 67
SARCASM AS A CONTRADICTION BETWEEN A
TWEET AND ITS TEMPORAL FACTS: A PATTERN-
BASED APPROACH
Santosh Kumar Bharti and Korra Sathya Babu
Department of Computer Science and Engineering, National Institute of Technology,
Rourkela, Odisha, India
ABSTRACT
In the context of Indian languages, sarcasm detection in Hindi is a tedious job as it is rich in morphology
and complex in structure. The annotated resources for sarcastic Hindi sentences are almost negligible for
machine learning analysis. Here, we propose a pattern-based framework for sarcasm detection in Hindi
tweets. It has been observed that a tweet is sarcastic if it contradicts its temporal facts intentionally. The
temporal fact is a collection of time-dependent facts which may change over the period. We used Hindi
news with timestamp as a corpus of temporal facts. The timestamp describes the fact period of any entity. In
this research, a temporal fact is represented as a <key, value> pair. To form a <key, value> pair, one need
to extract triplets i.e. subject, verb and object for every sentence. Next, a key is formed using subject and
verb together. The value is formed using object and timestamp together. To predict the sarcastic tweet; one
needs to extract the triplets from input tweet and form a <key, value> pair. Now, the <key, value> pair of
the input tweet is mapped with related <key, value> pair in the corpus of temporal facts and are checked if
they coincide. If they contradict, the input tweet is considered as sarcastic. The achieved accuracy of the
proposed approach outperforms the state-of-the-arts techniques for Hindi sarcasm detection as it attains
an accuracy of 82.8%.
KEYWORDS
Natural Language Processing, Sentiment, Sarcasm, Social Network, Tweets
1. INTRODUCTION
Sarcasm is an indirect way to negate a sentence to feel someone insult or stupid. While writing or
speaking sarcastic sentences, one often use positive words or intensified positive words.
Therefore, detection of sarcasm is one of the interesting problems in the text, speech, video, etc.
In textual data, physical clues are missing such as hands movement, rolling eyes, intentional tonal
stress, etc., unlike speech and video data. This makes the analysis of sarcasm in text ever more
difficult.
Twitter is one of the popular social media sites that allow users to post a message within 140
characters. The data generated by Twitter is much higher compared to other social media [1]. For
instance, Twitter claims to have more than 500 million users, out of which more than 332 million
are active [2]. They post more than 340 million tweets and 1.6 billion search queries every day
[1]. Due to this, Twitter generates such a large volume of data every day that poses many
challenges such as accessing, storing, processing, analyzing etc. [3]. Among these challenges,
analysis of sarcasm in tweets is the most difficult task especially in the context of Indian language
such as Hindi, Telugu, Bengali, etc.

68
Hindi is one of the popular Indian languages and is highly used by Indians while communicating
on social media such as Facebook, Twitter, WhatsApp, etc. Hindi stands fourth in popularity after
Mandarin, Spanish, and English [4]. It is widely used for speaking in countries like India,
Mauritius, Fiji, Suriname, Guyana, Trinidad & Tobago and Nepal [5]. It has 490 million speakers
across the world, and majority of them are from India [6]. In the past, several authors have
proposed sarcasm detection system for tweets scripted in English [7-15]. These methods require
an extensive set of annotated training data to detect sarcasm. In the domain of sarcasm detection,
availability of Hindi annotated dataset is almost negligible. Due to this, the research related to the
Hindi language is not much explored so far. A sarcasm detection system was developed [16] for
Hindi tweets using a similar set of features used for English tweets namely, #tag, emoticons,
punctuation marks etc. However, in real scenario natural Hindi tweets 1 are different in structure
unlike English scripted tweets or Hindi tweets translated from English tweets. A sample list of
Hindi sarcastic tweets is shown in Figure 1.
Figure 1: Sample Hindi sarcastic tweets.
To identify sarcasm in such natural Hindi tweets, the same feature set used for English scripted
tweets might not be applicable efficiently. Therefore, one needs to rely on other parameters such
as news context, specific patterns, rules, etc., to identify sarcastic Hindi tweets. In this research,
we propose a pattern-based framework for sarcasm detection in Hindi tweets by observing
manually annotated set of natural Hindi sarcastic tweets. After analyzing these tweets, it is
observed that a tweet is sarcastic if it contradicts to its temporal fact. A temporal fact can be
defined as a time-dependent fact. Unlike a universal fact, the temporal fact may change over the
period. This approach is wholly based on timestamp that covers the facts period. The news is one
of the primary sources of temporal facts over the period. Hence, this research uses a corpus of
Hindi news from top Hindi news sources as a corpus of temporal facts. While prediction, the
framework verifies the occurrences of any contradiction between an input tweet and its temporal
fact. If they contradict, the input tweet is considered as sarcastic; otherwise not sarcastic.
The rest of the paper is organised as follows: Section 2 describes related work. All the
preliminaries are explained in Section 3. The proposed scheme is discussed in Section 4. Analysis
of the results are given in Section 5 and conclusion of the article is drawn in Section 6.
2. RELATED WORK
In recent past, several authors have worked on sarcasm detection in Twitter data. They proposed
various sarcasm detection systems for tweets scripted in English [7-15]. According to the
literature, Twitter sarcasm detection techniques can be classified into four categories, namely
lexicon-based approach, machine learning-based approach, context-based approach and pattern-
based approach.

69
2.1 LEXICON-BASED APPROACH
This approach utilizes a bags-of-lexicons which comprise of unigram, bigram, trigram, etc.
phrases to identify sarcasm in tweets. Riloff et al. [9] developed two bags-of-lexicons, namely
positive sentiment and negative situation using bootstrapping technique. These lexicon file
consists of unigram, bigram, and trigram phrases. Further, they utilized these phrases to identify
sarcasm in tweets for the structure, positive sentiment in a negative situation. Similarly, Bharti et
al. [12] developed four bag-of-lexicons, namely positive sentiment, negative situation, negative
sentiment and positive situation using parsing technique. Further, they utilized these phrases to
identify sarcasm in tweets for the both structures, namely positive sentiment in a negative
situation and negative sentiment in a positive situation.
2.2 MACHINE LEARNING-BASED APPROACH
In this approach, one needs to get labelled dataset to extract features. Further, these features are
used to train the classifiers such as SVM, DT, NB, etc. Liebrecht et al. [8] designed a machine
learning model to detect sarcasm in Dutch tweets. Peng et al. [17] used a supervised approach to
classify movie reviews into two classes after performing subjective feature extractions. Tayal et
al. [18] utilized machine learning classifier to detect the sarcastic political tweets. Tungthamthiti
et al. [19], explored concept level knowledge using the hyperbolic words in sentences and gave
an indirect contradiction between sentiment and situation such as raining, bad weather that are
conceptually the same. Therefore, if ‘raining’ is present in any sentence, then one can assume
‘bad weather’. They build a system for sarcasm detection using SVM classifier and used these
indirect contradiction feature for training. Bharti et al. [20] developed a machine learning
framework to detect sarcastic tweets. They deployed several classical classifiers such as SVM,
DT, NB, Maximum Entropy (ME), etc.
2.3 CONTEXT-BASED APPROACH
The relationship between an author and audience followed by the immediate communicative
context can be helpful to improve the sarcasm prediction accuracy [21]. Message-level sarcasm
detection on Twitter using a context-based model were used for sarcasm detection [22]. Chains of
tweets that work in a context were considered. They introduce a complex classification model that
works over an entire tweet sequence and not on one tweet at a time. Integration between linguistic
and contextual features extracted from the analysis of visuals embedded in multimodal posts was
deployed for sarcasm detection [23]. A framework based on the linguistic theory of context
incongruity and an introduction of inter-sentential incongruity by considering the history of the
posts in the discussion thread was considered for sarcasm detection [11]. A quantitative evidence
of historical tweets of an author can provide additional context for sarcasm detection [24]. The
author`s past sentiment on the entities in a tweet was exploited to detect the sarcastic intent.
2.4 PATTERN-BASED APPROACH
In this approach, one can observe the sarcastic dataset and identify unique patterns for sarcastic
text. Riloff et al. identified a unique pattern for sarcasm detection in the tweet i.e. “sarcasm is a
contrast between positive sentiment and negative situation” [9]. Similarly, another pattern-based
sarcasm detection system using parsing technique was proposed [12]. They generalised the
concept of [9] approach and suggested that sarcasm is a contradiction between sentiment and
situation. It may follow both the patterns namely, sarcasm is a contrast between positive
sentiment and negative situation, sarcasm is a contrast between negative sentiment and positive
situation. Bouazizi and Ohtsuki identified three patterns for sarcasm in tweets; namely, sarcasm is

70
a wit, sarcasm is an evasion and sarcasm is a whimper [14]. Based on these three patterns, they
proposed an efficient sarcasm detection system that enhance the accuracy of sentiment analysis.
In the context of Indian languages, sarcasm detection is less explored due to the unavailability of
benchmark resources for training and testing. Desai et al. proposed a Support Vector Machine
(SVM) based sarcasm detector for Hindi sentences [16]. They used Hindi tweets as the dataset for
training and testing using SVM classifier. In the absence of annotated datasets for training and
testing, they converted English tweets into Hindi. Therefore, they focused on a similar set of
features like emoticons and punctuation marks for sarcasm detection in English text. These
methods when applied directly, are not suitable for the Hindi sarcastic tweets. Similarly, a
context-based approach for sarcasm detection in Hindi tweets is proposed [25]. They exploit
online Hindi news as a context to determine the sarcasm in tweets. They experimented using a
small set of manually collected and annotated Hindi tweets. Dalal et al. proposed an insult
detection in Hindi text using logistic regression and SVM classifiers [26]. They have used n-
grams, negation and second person as features to build the feature vector to train these classifiers.
3. PRELIMINARIES
This section describes all the tools used in this article namely, Hindi POS tagging, Hindi parsing
and triplet extraction of a sentence. The details are explained below:
3.1 POS TAGGING
To identify the POS tag information in Hindi sentences, we have developed a Hidden Markov
Model (HMM) based POS tagger [30]. It uses Indian Language (IL) standard tagset which
consists of 24 tags [27]. The POS tagger is available on URL:http//www.taghindi.herokuapp.com.
The implementation details are released on URL: https://github.com/rkp768/hindi-pos-tagger. An
example Hindi POS tagging is shown in Figure 2.
Figure 2: An example POS tagging in Hindi text.
3.2 PARSING
Parsing is the way toward breaking down the linguistic structure of a language. Given a
succession of words, a parser shapes units like subject, verb, object and identify the relations
between these units, as per the tenets of a formal sentence structure and produce a parse tree
(Bharti, 2015). This article deploys Hindi shallow parser to get the parsing tree information for
Hindi text [28]. The Hindi shallow parser is available on
URL:http://ltrc.iiit.ac.in/analyzer/hindi/run.cgi. Figure 3 depict the examples of the parsing
information for Hindi sentence राम के ला खाता ह| (Ram kela khata hain) (Ram eats banana).

71
Figure 3: Parsing information for Hindi text ”राम के ला खाता ह”| (Ram kela khata hain) (Ram eats banana)
3.3 TRIPLET EXTRACTION
This article used Rusu_Triplets extraction algorithm [29] to extract the triplets i.e. subject, object
and verb of any sentence using parse tree information. For example, triplets information of the
Hindi sentence “राम के ला खाता ह”| (Ram kela khata hain) (Ram eats banana) are राम (Ram) as
subject, के ला (Kela) as object and खाता (Khata) as verb.
4. PROPOSED SCHEME
This section describes the proposed framework for sarcasm detection in Hindi tweets using its
temporal fact. In this research, a corpus of Hindi news is considered as a set of temporal facts and
prediction of sarcasm in Hindi tweet is made through a pattern “sarcasm as a contradiction
between a tweet and its temporal facts”. While detecting a sarcastic tweet, it checks the
contradiction between a tweet and its temporal fact. If contradiction occurs, then the tweet is
classified as sarcastic; otherwise non-sarcastic. The pipeline procedure of proposed framework is
shown in Figure 4.
The framework starts with a Hindi news corpus and identifies the parsing (syntax tree)
information of every news in the corpus. Next, the parse information is used to extract triplets i.e.
subject, object and verb of every news sentence. Based on the extracted triplets, possible tweets
are collected from Twitter. For testing a tweet as sarcastic or not, it forms (key, value) pairs for
both tweet and the related news. While forming (key, value) pair, subject and verb together from
triplets are treated as key. Object and its timestamp together are considered as value. Finally, it
checks the contradiction between a tweet and its temporal fact. If they contradict each other, the
input tweet is considered as sarcastic in the context of time-dependent fact; otherwise, given tweet
is not sarcastic.

72
Figure 4: Proposed framework for sarcasm detection in Hindi tweets using temporal facts.
4.1 NEWS AND TWEETS COLLECTION
This After browsing several online news sources such as ABP News Hindi, AajTak, Zee News
Hindi, etc., on Twitter, we have collected a total of around 500 processed Hindi news with times-
tamp manually using the trending topics such as #EVMs, #EVMtampering, #Bahubali2,
#khelratna, etc. A sample Hindi news on EVMs is given in Figure 5. While pre-processing, the
news related to murder, rape, bomb blast, etc., were eliminated. We believe that sarcastic tweets
will not be floated on such grave topics. Using the similar trending topics, approximately 2500
Hindi tweets were collected manually in the same timestamp. A sample of tweets on #EVMs is
shown in Figure 6. The datasets of both the news and tweets are released on URL:
https://github.com/sbharti1984/Hindi-Tweets.
Figure 5: A sample Hindi news on EVM.
Further, among 2500 Hindi tweets, 2000 Hindi tweets were used as observation set and remaining
500 Hindi tweets were used as testing set. Next, both sets of Hindi tweets were distributed for
annotation among three professionals in the Hindi language who are teachers and practitioners.
They annotated these 2500 tweets manually to identify the tweet as sarcastic or not. After
collecting the annotation results from all the individuals and we calculate the Inter-annotator
agreement (IAA). To compute the IAA, Fleiss’ Kappa coefficient [31] is used as it is more
suitable when the number of annotators is more than two. In this experiment, the attained IAA
value is 0.89.

73
Figure 6: A sample Hindi tweets related to EVM.
The annotation result of testing set is used as ground truth to check the performance of the pro-
posed system and error analysis. After verifying the annotation results of all the three annotators,
the result is shown in Table 1.
Table 1: Annotation result of 2500 Hindi tweets as sarcastic or not.
Total tweets sarcastic not sarcastic ambiguous
Observation set 2000 724 1140 136
Testing set 500 184 290 26
After annotation, we observed that most of the sarcastic tweets contradict the temporal facts.
Therefore, we proposed a pattern-based sarcasm detection system using pattern “sarcasm as a
contradiction between a tweet and its temporal facts”.
4.2 GENERATION OF (KEY, VALUE) PAIRS FROM TEMPORAL FACTS (NEWS)
This section generates the <key, value> pairs of news corpus as a set of temporal facts. The
procedure of generating <key, value> pairs is given in Algorithm 1.
Algorithm 1 takes a corpus of news as temporal facts and finds parse tree information of every
news using Hindi shallow parser [28]. Next, the parse tree information is used to extract triplets
(subject, verb and object) using Rusu_Triplets extraction algorithm [29]. Finally, it generates the
(key, value) pair for every news sentence with the help of the extracted triplets. In the formation
of (key, value) pair of news, a key is made up using subject and verb from its triplets. Value is
consists of an object and its timestamp. The timestamp is the date of which the news was
released.

74
4.3 SARCASM DETECTION ALGORITHM
This section identifies a sarcastic tweet based on the contradiction between a tweet and its
temporal facts (TF). In this approach, a news corpus is used as the set of temporal facts. The
procedure of identifying sarcastic tweets is given in Algorithm 2.

75
Algorithm 2 takes tweets corpus and temporal fact file (TFF) as the input and finds the parse tree
information of every testing tweet. Next, it ex- tracts triplets of every tweet using parse
information and generates the <key, value> pair. Further, it compares the <key, value> pair of a
testing tweet with one of the <key, value> pair in temporal fact file to identify sarcastic tweet. If
both the key and value of both tweet and temporal facts are matching, then the tweet is not
sarcastic. Otherwise, tweets are sarcastic.
Let us consider an example shown in Figure 7 to test if a tweet is sarcastic or not. The example
consists of a tweet and corresponding related news. The <key, value> pair of given tweet in the
example is <(Sourav Bhardwaj, dabaya), (EVM secret code, 10 May)>. The temporal fact from
the related news is “there is no secret code in EVMs”. The <key, value> pair of related news is
<(EVMs, istemal nahi), (secret code, 10 May)>. Hence, given tweet in the example is sarcastic as
it contradicts its temporal facts.
Figure 7: An example of input tweet and related news based on #EVM.
Similarly, let us consider another example of input tweet and related news based on
#khelratna as shown in Figure 8. According to proposed algorithm, the <key, value> pair
of the given tweet in the example is <(Devendra Jhajharia, Hasil Karne), (KhelRatna, 03
Aug)>. The temporal fact from the related news is “Devendra Jhajharia and hockey player Sardar
Singh will get KhelRatna”. The <key, value> pair of related news is <(Devendra Jhajharia and
Hockey Khiladi Sardar Singh, Ko Milega), (KhelRatna, 03 Aug)>. Hence, given tweet in the
example is not sarcastic as tweet infer similar meaning as temporal facts. Here, the meaning of
“Hasil Karna” and “Milega” is semantically same in this context of tweet and news.
Figure 8: Example of input tweet and related news based on #khelratna.
5. EXPERIMENTAL RESULTS AND DISCUSSION
This section describes the experimental results of the proposed approach to identify sarcasm in
Hindi tweets. There are four statistical parameters considered namely, accuracy, precision, recall
and F1-measure to evaluate the performance of proposed approach. The accuracy determines the
performance of a system. A formula to calculate accuracy is given in Equation 1.

76
Accuracy = (Tp +Tn)/ (Tp +Tn +Fp +Fn) (1)
Precision and recall are the other parameters which determine the performance of a system. Here,
precision determines how much relevant information system identified. Whereas, recall
determines how much identified information is relevant. The formula to ascertain precision and
recall appeared in equations 2and 3.
Precision = Tp/ (Tp + Fp) (2)
Recall = Tp/ (Tp + Fn) (3)
The value of precision and recall may vary application to application. For example, an
application attains high precision but low recall. Similarly, another application attains low
precision but high recall. To deal with this situation, one can rely on another statistical
parameter i.e. F1- measure. It is a harmonic mean of precision and recall. To obtain the F1-
measure, a formula is given in Equation 4.
F1−measure = 2∗Precision∗ Recall
Precision + Recall (4)
Where,
Tp = true positive, Fp = false positive, Fn = false negative.
5.1 EXPERIMENTAL RESULTS
A set of 500 random Hindi tweets is used as a testing set to experiment. After the experiment, a
confusion matrix of 500 Hindi tweets is given in Table 2 for error analysis. To identify sarcasm in
tweets, the proposed system identified 161 tweets correctly out of 184 sarcastic tweets. The
ground truth of sarcastic and non-sarcastic tweets are given in Table 1. Using the confusion
matrix given in Table 2, the values of precision, recall, F1-measure and accuracy, attained by the
proposed approach for identifying sarcasm in tweets are given in Table 3.
Table 2: Confusion matrix for sarcasm detection in Hindi tweets.
Proposed approach No. of tweets Tp Tn Fp Fn
Identifying sarcasm 500 161 253 39 47
Table 3: Accuracy, Precision, Recall and F1-measure.
Proposed approach Precision Recall F1-measure Accuracy (%)
Identifying sarcasm 0.805 0.774 0.786 82.8
We made a comparison of the proposed approach with state-of-the-art approaches and is shown in
Table 4.

77
Table 4: Comparison of proposed approach with some of the state-of-the-art techniques.
Study Precision Recall F1-measure Accuracy (%)
Desai et al. 0.732 0.674 0.705 71.4
Bharti et al. 0.736 0.717 0.726 79.4
Proposed Approach 0.834 0.831 0.832 82.4
5.2 DISCUSSION
While identifying sarcasm of Hindi tweets in the context of temporal facts, the algorithm checks
for the contradiction between input tweet and the temporal fact obtained from related news in the
same timestamp. If the user posts any tweet which negates the temporal fact intentionally on the
same timestamp, then the input tweet is sarcastic. Here, another example is discussed and shown
in Figure 9 in support of proposed approach.
Figure 9: Another example of input tweet and related news based on #EVM.
The example consists of a tweet and corresponding related news. The <key, value> pair of the
given tweet in the example is <(Kalyugi Yudhishthira, hack), (EVM, 10 May)>. The temporal
fact from the related news is “EVMs cannot be hacked”. The <key, value> pair of related news is
<(AAP ka Dava, kharij karna), (EVM hack, 10 May)>. Hence, given tweet in the example is
sarcastic as it contradicts its temporal facts.
6. CONCLUSIONS AND FUTURE DIRECTION
In the absence of sufficient annotated sarcastic Hindi tweets for training and testing, a set of
sarcastic Hindi tweets were collected and annotated manually. After observing the sarcastic Hindi
tweets, we found that most of the sarcastic tweets were written intentionally and it contradicts the
time-dependent fact of the same timestamp. Therefore, this article proposed a pattern-based
approach to identify sarcastic Hindi tweets using temporal facts and timestamp as the context.
Here, a set of online news is treated as temporal facts. The proposed approach attains an accuracy
of 82.8%. The performance of the proposed approach is compared with state-of-the-arts
techniques for Hindi sarcasm detection and observe that proposed approach outperforms existing
approaches. In future, we will use some new #tag trending to collect a large set of tweets as
observation set and formulate more pattern for sarcastic tweets in Hindi.
REFERENCES
[1] Chaffey, D. 2016. Global social media research summary 2016. URL
http://www.smartinsights.com/social-media-marketing/social-media-strategy/new-globalsocial-media-
research.
[2] Tan, W., Blake, M. B., Saleh, I. and Dustdar, S. 2013. Social-network sourced big data analytics.
Internet Computing, 17(5), pp. 62-69.
[3] Gastelum, Z. N., Whattam, K. M. 2013. State-of-the- Art of Social Media Analytics Research. Pacific
Northwest National Laboratory (PNNL), pp. 1-9.

78
[4] Lee, Parkvall, M. 2007. Varldens 100 storsta sprak 2007. The Worlds 100, 2007.
[5] Mesthrie, R. 1992. Language in indenture: a sociolinguistic history of Bhojpuri-Hindi. South Africa,
Routledge, pp. 30-32. ISBN 978-0415064040.
[6] Language and Culture. 2015. Top 30 Languages by Number of Native Speakers, 2005.
http://www.vistawide.com/languages/top 30 languages.htm.
[7] Gonzalez-Ibanez, R., Muresan, S. and Wacholder, N. 2011. Identifying sarcasm in twitter: a closer
look. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics:
Human Language Technologies, vol. 2, pp. 581-586.
[8] Liebrecht, C. Kunneman, F. and van den Bosch, A. 2013. The perfect solution for detecting sarcasm
in tweets #not. In proceedings of the Association for Computational Linguistics, pp. 29-37.
[9] Riloff, E., Qadir, A., Surve, P., De Silva, L., Gilbert, N. and Huang, R. 2013. Sarcasm as contrast
between a positive sentiment and negative situation. In proceedings of the Empirical methods in
natural language processing, pp. 704-714.
[10] Rajadesingan, A., Zafarani, R. and Liu, H. 2015. Sarcasm detection on Twitter: A behavioural
modelling approach. In Proceedings of the Eighth ACM Inter- national Conference on Web Search
and Data Mining, pp. 97-106.
[11] Joshi, A., Sharma, V. and Bhattacharyya, P. 2015. Harnessing Context Incongruity for Sarcasm
Detection. In proceedings of the Association for Computational Linguistics vol. 2, pp. 757-762.
[12] Bharti, S., Sathya Babu, K. and Jena, S. 2015. Parsing- based sarcasm sentiment recognition in
Twitter data. In International Conference on Advances in Social Networks Analysis and Mining
(ASONAM), pp. 1373-1380, IEEE/ACM.
[13] Joshi, A., Tripathi, V., Patel, K., Bhattacharyya, P. and Carman, M. 2016. Are Word Embedding-
based Features Useful for Sarcasm Detection?. ArXiv preprint arXiv: 1610.00883.
[14] Bouazizi, M. and Ohtsuki, T. O. 2016. A pattern-based approach for sarcasm detection on twitter.
IEEE Access, vol. 4, pp. 5477-5488.
[15] Bharti, S.K., Vachha, B. Pradhan, R.K., Babu, K.S. and Jena, S.K. 2016. Sarcastic sentiment detection
in tweets streamed in real time: a big data approach. Digital Communications and Networks 2(3), pp.
108-121.
[16] Desai, N. and Dave, A.D. 2016. Sarcasm Detection in Hindi sentences using Support Vector machine.
International Journal of Advance Research in Computer Science and Management Studies, 4(7), pp.
8- 15.
[17] P. Liu, W. Chen, G. Ou, T. Wang, D. Yang, and K. Lei. 2014. Sarcasm detection in social media
based on imbalanced classification. In Web Age Information Management, pp. 459–471.
[18] D. Tayal, S. Yadav, K. Gupta, B. Rajput, and K. Kumari. 2014. Polarity detection of sarcastic
political tweets. In proceedings of International Conference on Computing for Sustainable Global
Development (INDIACom), pp. 625–628, IEEE.
[19] P. Tungthamthiti, K. Shirai, and M. Mohd. 2014. Recognition of sarcasm in tweets based on concept
level sentiment analysis and supervised learning approaches. ACL, pp. 404–413.
[20] S. K. Bharti, R. Pradhan, K. S. Babu, and S. K. Jena. 2017. Sarcasm analysis on twitter data using
machine learning approaches. In Trends in Social Network Analysis, pp. 51–76, Springer.
[21] D. Bamman and N. A. Smith. 2015. Contextualized sarcasm detection on Twitter. In ICWSM, pp.
574–577.
[22] Z. Wang, Z. Wu, R. Wang, and Y. Ren. 2015. Twitter sarcasm detection exploiting a context-based
model. In International Conference on Web Information Systems Engineering, pp. 77–91, Springer.
[23] R. Schifanella, P. de Juan, J. Tetreault, and L. Cao. 2016. Detecting sarcasm in multimodal social
platforms. In Proceedings of the 2016 ACM on Multimedia Conference, pp. 1136–1145, ACM.
[24] A. Khattri, A. Joshi, P. Bhattacharyya, and M. J. Carman. 2015. Your sentiment precedes you: Using
an author’s historical tweets to predict sarcasm. In Proceedings of the WASSA@ EMNLP, pp. 25–30.
[25] Bharti, S., Sathya Babu, K. and Jena, S. 2017. Harnessing Online News for Sarcasm Detection in
Hindi Tweets. Proceeding of PReMI 2017, LNCS, Springer.
[26] Dalal, C., Tandon, S. and Mukerjee, A. 2014. Insult Detection in Hindi. Technical report on Artificial
Intelligence, pp. 1-8.
[27] Bharati, A., Sangal, R., Sharma, D. and Bai, L. 2006. Anncorra: Annotating corpora guidelines for
POS and chunk annotation for Indian languages. LTRC- TR31, pp. 1-38.
[28] Bharati, A. and Mannem, P.R. 2007. Introduction to shallow parsing contest on south Asian
languages. In Proceedings of the IJCAI and the Workshop On Shallow Parsing for South Asian
Languages (SPSAL), pp. 1-8.

79
[29] Rusu, D., Dali, L., Fortuna, B., Grobelnik, M. and D. Mladenic. 2007. Triplet extraction from
sentences. In Proceedings of the 10th International Multiconference on Information Society-IS, pp. 8-
12.
[30] Doug, et al. 1992. A practical part-of-speech tagger. In Proceedings of the third conference on applied
natural language processing. Association for Computational Linguistics, pp. 133-140.
[31] Fleiss, J.L., Cohen, J. and Everitt, B.S. 1969. Large sample standard errors of kappa and weighted
kappa. In Psychological Bulletin. American Psychological Association, 72(5), pp. 323-
AUTHORS
Santosh Kumar Bharti is currently pursuing his Ph.D. in CSE from National
Institute of Technology Rourkela, India. His research interest includes opinion mining
and sarcasm sentiment detection resume.
Korra Sathya Babu is working as an Assistant Professor in the Dept. of CSE,
National Institute of Technology Rourkela India. His research interest includes Data
engineering, Data privacy, Opinion mining and Sarcasm sentiment detection.

SARCASM AS A CONTRADICTION BETWEEN A TWEET AND ITS TEMPORAL FACTS: A PATTERNBASED APPROACH

More Related Content

What's hot (11)

Similar to SARCASM AS A CONTRADICTION BETWEEN A TWEET AND ITS TEMPORAL FACTS: A PATTERNBASED APPROACH (20)

More from kevig (20)

Recently uploaded (20)

SARCASM AS A CONTRADICTION BETWEEN A TWEET AND ITS TEMPORAL FACTS: A PATTERNBASED APPROACH