Detecting Sarcasm is Extremely Easy ;-) Page: 22
6 p.View a full description of this paper.
Extracted Text
The following text was automatically extracted from the image on this page using optical character recognition software:
Maynard and Greenwood (2014) learned hashtags
that commonly correspond with sarcastic tweets,
and checked for those in subsequent tweets to de-
termine whether or not the tweets were sarcastic.
Other researchers utilized Twitter histories, devel-
oping behavioral models of sarcasm usage specific
to individual users (Rajadesingan et al., 2015), or
features based on the users, their audiences, and
the author-audience relationship of the tweet in
question (Bamman and Smith, 2015). Some re-
searchers considered the sentiment (Riloff et al.,
2013) or emotional scenario (Reyes et al., 2013) of
a tweet when deciding whether or not it contained
sarcasm, and finally others experimented with n-
grams (Liebrecht et al., 2013) and word embed-
dings (Ghosh et al., 2015; Ghosh and Veale, 2016;
Amir et al., 2016).
Amazon product reviews, which have also in-
terested sarcasm researchers, differ from tweets in
several key ways: they are of variable (and of-
ten much longer) length, they do not utilize hash-
tags, and they generally contain more context. The
primary domain-specific feature employed by sar-
casm detection researchers using Amazon prod-
uct reviews has been a product's "star rating" (the
number of stars assigned to the product by the
review writer) (Buschmeier et al., 2014; Parde
and Nielsen, 2017). Other characteristics that re-
searchers have considered in this domain include
syntactic features (Buschmeier et al., 2014; Davi-
dov et al., 2010) and the presence of interjections
or laughter terms (Buschmeier et al., 2014).
Finally, we learned a general sarcasm detection
model from many tweets and fewer Amazon prod-
uct reviews (Parde and Nielsen, 2017). We found
that by applying a domain adaptation step prior
to training the model, we were able to achieve
higher performance in predicting sarcasm in Ama-
zon product reviews over models that trained on
reviews alone or on a simple combination of re-
views and tweets. Our prior work was notable
in that it was the first approach that specifically
sought domain-generality. We analyze its perfor-
mance on different datasets in this work.
3 Sarcasm Detection Methods
We train our sarcasm detection approach on the
same training data used in our previous work(3998 tweets and 1003 Amazon product reviews),
and apply it to two test datasets: AMAZON,
a 251-instance set of sarcastic (87) and non-sarcastic (164) Amazon product reviews origi-
nally collected by Filatova (2012), and TWIT-
TER, a 1000-instance set of sarcastic (391) and
non-sarcastic (609) tweets containing the hash-
tags #sarcasm (the sarcastic class) or #happiness,
#sadness, #anger, #surprise, #fear, or #disgust
(the negative class).' The approach utilizes fea-
tures that seek to convey informative characteris-
tics from the domains considered as well as gen-
eral characteristics expected to remain indicative
of sarcasm across many domains. We briefly de-
scribe each in Table 1; for additional information,
the reader is referred to our earlier paper.
3.1 Classification Algorithm
All features were extracted from each instance, re-
gardless of its domain (feature values were left
empty when it was impossible to fill them, e.g.,
star rating for tweets). Then, the feature space
was transformed using the domain adaptation ap-
proach originally outlined by Daum6 III (2007).
Daum6's approach works by modifying the fea-
ture space such that it contains three mappings of
the original features: a source version, a target ver-
sion, and a general version. More formally, letting
X = R3F be the augmented version of a feature
space X = RF, and V, t : X -+ X be map-
pings for the source and target data, respectively,
V (X) = (x, 0, x), -t (x) = (0, x, x) (1)
where 0 (0, 0, ..., 0) E RF is the zero vector. It
is then left to the classification algorithm to decide
how to best take advantage of this supplemental
information. We use Nafve Bayes, following our
earlier work.
4 Model Performance
We compute precision (P), recall (R), and f-
measure (F) on the positive (sarcastic) class for
both TWITTER and AMAZON, and report results
relative to the performance of other systems on the
same data (Table 2). Our results on AMAZON are
identical to those reported originally (Parde and
Nielsen, 2017). Our previous paper reported re-
sults on TWITTER when training only on Twitter
data; here we instead apply the same model as ap-
plied to AMAZON and achieve slightly higher re-sults. Thus, the approach outperforms other sar-
'These hashtags were removed prior to using the data.22
Upcoming Pages
Here’s what’s next.
Search Inside
This paper can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Paper.
Parde, Natalie & Nielsen, Rodney D. Detecting Sarcasm is Extremely Easy ;-), paper, June 5, 2018; Stroudsburg, Pennsylvania. (https://digital.library.unt.edu/ark:/67531/metadc1164535/m1/2/: accessed July 18, 2024), University of North Texas Libraries, UNT Digital Library, https://digital.library.unt.edu; crediting UNT College of Engineering.