Citation Context

... analysis in the phrasal and sentence level, see (Yu and Hatzivassiloglou, 2003; Wilson et al., 2005; McDonald et al., 2007; Titov and McDonald, 2008a; Titov and McDonald, 2008b; Wilson et al., 2009; =-=Tsur et al., 2010-=-) among others. Only a few studies analyze the sentiment and polarity of tweets targeted at major brands. Jansen et al. (2009) used a commercial sentiment analyzer as well as a manually labeled corpus...

Citation Context

...r and Rappoport, 2009). All systems currently fail to correctly classify the sentiment of sarcastic sentences. In this paper we utilize the semi-supervised sarcasm identification algorithm (SASI) of (=-=Tsur et al., 2010-=-). The algorithm employs two modules: semi supervised pattern acquisition for identifying sarcastic patterns that serve as features for a classifier, and a classification stage that classifies each se...

Citation Context

... Table 2. The sarcasm detection performed very well at 91% Precision and Recall. Out 6http://www.trend-miner.eu 4241 of 400 tweets, there were 91 sarcastic sentences, which is quite a high proportion, and many of these were not indicated by any kind of sarcasm marker. Twitter corpus P R F1 Opinionated 65.69 77.31 71.02 Opinion+polarity 52.61 61.92 56.89 Polarity only 80.08 80.03 80.05 Sarcasm detection 91.03 91.04 91.03 Table 2: Experiments on General Tweets 6. Related Work There have been a few recent works attempting to detect sarcasm in tweets and other user-generated content. Tsur et al. (Tsur et al., 2010) use a semi-supervised approach to classify sentences in online product reviews into different sarcastic classes, and report an F-measure of 82.7% on the binary sarcasm detection task (although Precision is much higher than Recall). Liebrecht et al. (Liebrecht et al., 2013) use the Balanced Winnow Algorithm to classify Dutch tweets as sarcastic or not, with 75% accuracy, training over a set of tweets with the #sarcasm hashtag. Reyes et al. (Reyes et al., 2013) use a similar technique on English tweets to detect ironic tweets, using the #irony hashtag, with 70% accuracy. Davidov et al. (Davidov...

Citation Context

... sarcasm detection, however. Book reviews are one example due to the repeated use of stock sarcastic types of phrase, such as “this book has a great cover” that can be learned from a training corpus (=-=Tsur, Davidov, & Rappoport, 2010-=-). Sarcasm in Portuguese political discussions can also be identified through a combination of features including the use of a politician’s name in diminutive form (Carvalho, Sarmento, Silva, & de Oli...

Citation Context

...sence is a latent construct and not clearly observable, we based our work on the previous work that focused also on mining latent constructs. The work done on opinion mining of online product reviews =-=[3, 15, 23]-=-, gender style differences [13] and sentiment analysis [4] are some of the main areas of research that informed our classification approach. The text classification tasks have been studied in the cont...

Citation Context

...et al., 2009), detection of synonyms (Turney, 2008a), disambiguation of nominal compound relations (Davidov and Rappoport, 2008a), sentiment analysis (Davidov et al., 2010b) and detection of sarcasm (=-=Tsur et al., 2010-=-). 9 Conclusion The main goal of this paper is to measure to what extent authors of micro-messages can be identified. We have shown that authors of very short texts can be successfully identified in a...

Citation Context

...s dependency parse tuples (e.g. [4]). However, such features require running expensive parsing models during the evaluation phase. Another approach is to hand-craft regular expressions, which require human effort but are cheaper to apply; they can also be used to implement business decisions or system constraints, which could be difficult to model in a statistical system. A method which merges the benefits of these two approaches involves learning phrase patterns, extensions of n-grams allowing gaps between words. Such features have been used successfully in a variety of text processing tasks [6, 7, 8, 9, 10]. In this paper, we propose a novel method of learning phrase patterns. Section 2 describes the core phrase pattern learning algorithm and contrasts it with previous work. The experimental setup and results are described in section 3. We discuss additional experiments aimed at system analysis in section 4 and conclude with suggestions for future work in section 5. 2. Phrase Pattern Learning Given a word sequence [wn1 ], we define a phrase pattern as a subsequence [wi1 , . . . , wik ] with 1 ≤ i1 < . . . ik ≤ n. The words wik for each pattern are referred to here as the “pivots” of the pattern....