Homographic Puns Recognition Based on Latent Semantic Structures

Computational Humor Recognition: A Systematic Literature Review

“…This linguistic phenomenon that occurs when a word or a phrase can express multiple meanings simultaneously is called ambiguity. It is the most common used feature found in our research (Kao et al 2015;Ortega-Bueno et al 2018;Reyes et al 2009a, b;Zaveri 2017, 2020;Sjöbergh and Araki 2007;Zhang et al 2017;Liu et al 2018a, b;Kamal and Abulaish 2019;Barbieri and Saggion 2014;Attardo et al 1994;Bucaria 2004; van den Beukel and Aroyo 2018; Diao et al 2017Hasan et al 2021;Reyes et al 2012;Yang et al 2015;Raz 2012;Morales and Zhai 2017;Ermilov et al 2018;Zhang et al 2018). Several research papers have reported that ambiguity is useful in humor recognition (Morales and Zhai 2017;Reyes et al 2009a;Hasan et al 2021;Attardo et al 1994;Bucaria 2004; van den Beukel and Aroyo 2018; Yang et al 2015;Kao et al 2015).…”

“…Consequently, many researchers have used this feature for humor classi cation (Diao et al 2017Zhang and Liu 2014;Skalicky et al 2016;van den Beukel and Aroyo 2018;Shahaf et al 2015;Ren et al 2021;Ortega-Bueno et al 2018;Mahajan and Zaveri 2020;Reyes et al 2009aReyes et al , 2012Kamal and Abulaish 2019;Ermilov et al 2018;Bucaria 2004). Zhang and Liu (2014) in their work of humor recognition on Twitter designed ve POS ratios of features (ratios of nouns, pronouns, verbs, and modi ers) and three POS pattern chains of features (number and length of patterns like "noun-verb-adjective").…”

“…Several researchers have tried to identify the presence of this humor property using the WordNet lexical database antonyms for English words, among other relations (Mihalcea and Strapparava 2005;Castro et al 2016;Zhang et al 2018;Vadehra 2017;Diao et al 2017;van den Beukel and Aroyo 2018;Ortega-Bueno et al 2018). For example, Diao et al (2017) captured the antonymy relationship between words in a sentence (pun) using the WordNet lexical database and designing two features: the antonym existence and the antonyms largest/average number of words. However, Mihalcea and Strapparava (2005) reported that despite the relatively large number of antonyms in WordNet, its coverage is far from complete.…”

“…However, their report contrasts with the results described by the work of Mihalcea and Pulman (2007) where they used polarity to label words with negative orientations from one-liners dataset as potential re ections of the incongruity based theories of humor and suggested the relevance of negative information to generate humor. Furthermore, Diao et al (2017) in their effort to identify the word-level emotion and affect they used the open resource SenticNet which enables them to design polarity features such as: sum, average, total absolute and average absolute of the polarity scores for all the words.…”

“…Stylistic features represent typographical elements such as length of word, number of words, length of text, word length mean, conjunctions, emoticons, hashtags, urls, deictic, acronyms, interrogative pronouns, uppercase words, parenthetical words, emphasized words, question words, number of punctuation, exclamation and quotation marks. Many researchers exploited such features to reveal the stylistic characteristics of humor (Barbieri and Saggion 2014;Adams 2012;Castro et al 2016;Westbury and Hollis 2019;Diao et al 2017;Purandare and Litman 2006;Reyes et al 2009b;Ermilov et al 2018;Jensen et al 2020;Zhang and Liu 2014;Liu et al 2018a;Khandelwal et al 2018;Raz 2012;Zaveri 2017, 2020;Ortega-Bueno et al 2018). In his quantitative analysis of linguistic markers used to signal humor, Adams (2012) using the UAM corpus examined the frequencies of ve categories of features: punctuation (ellipsis, exclamation mark, and quotation marks not used for actual quoted speech), formatting (caps lock, elongation, and spelling variations used for effect), emoticons, laughter (textual and acronym) and explicit.…”