A Novel Annotation Schema for Conversational Humor: Capturing the

cultural nuances in Kanyasulkam

Vaishnavi Pamulapati Gayatri Purigilla Radhika Mamidi
Language Technologies Research Center
International Institute of Information Technology
Hyderabad, India
vaishnavi.p@research.iiit.ac.in
gayatri.purigilla@research.iiit.ac.in
radhika.mamidi@iiit.ac.in

Abstract

Humor research is a multifaceted field that has led to a better understanding of humor’s
psychological effects and the development of different theories of humor. This paper’s main
objective is to develop a hierarchical schema for a fine-grained annotation of Conversational
Humor. A prominent 19th century play from Telugu, Kanyasulkam, is annotated to substantiate
the work across cultures at multiple levels. Based on the Benign Violation Theory, the benignity
or non-benignity of the interlocutor’s intentions is included within the framework. Under the
categories mentioned above, in addition to different types of humor, the techniques utilized by
these types are identified. Furthermore, the inter-annotator agreement is calculated to assess the
accuracy and validity of the dataset. An in-depth analysis of the disagreement is performed to
understand the subjectivity of humor better.

Keywords:conversational humor, Benign Violation Theory, multicultural, hierarchical schema

1 Introduction
Humor and its dependence on society and culture have been the focus of research since times immemorial
(Raskin, 1979). From finding theories to define humor (Raskin, 1985; Meyer, 2000; Attardo and Raskin,
1991) to an analysis of the perception of humor in jokes (Raskin, 1979), humor studies have been proved
to be an essential aspect of linguistic as well as sociological, psychological and philosophical research.
Many papers discuss types of humor (Dynel, 2009; Alexander, 1997; Behrens, 1977), but this paper
stands apart. It focuses on creating an annotation schema for conversational humor with a stage play as
the medium of analysis while claiming that this schema can be used across languages. Conversational
humor is the spontaneous or pre-constructed interactional humor. The interlocutors intend to amuse the
listener directly or shift to a humorous frame where there is humor beyond what the literal verbalizations
convey (Dynel, 2009). Stage play is chosen as the medium of analysis since ’conversational humor’ is
an umbrella term that covers various semantic and pragmatic types of humor that occur in interpersonal
conversation, both real-life and fictional (Dynel, 2009).
There are key differences between plays and other forms of discourse, like transcribed recordings of
actual conversations or novels, justifying our use of stage play in the paper. The differences include but
are not limited to pauses, pause fillers, and discourse markers as essential features of characterization in
a play, unlike their use in actual conversations. In a play, there is more character-character interaction
than in novels, which have more narration from one point of view (Wareing and Thornborrow, 1998).
But the annotation schema presented here does not restrict its application on plays alone but can also
cover novels, TV shows, movies, etc; essentially any genre that involves a premeditated conversation.
This paper also focuses on how humor’s form and function are influenced multiculturally by annotating
one of the most famous plays of the Telugu culture, Kanyasulkam. Studies show that culture plays a
vital role in conversational humor in some distinct ways like the need for shared knowledge and standard
references, and others more indirect, like how the importance given to language awareness by any culture
This work is licensed under a Creative Commons Attribution 4.0 International License. License details: http://
creativecommons.org/licenses/by/4.0/.

34
The 14th Linguistic Annotation Workshop, pages 34–47
Barcelona, Spain (Online), December 12, 2020.
dictates the preference for wit and linguistic play (Mullan and Beal, 2018). Kanyasulkam is a play set in
the 19th century Vijayanagaram which uses humor to talk about the social evils prevalent in the society.
However, while the author talks about child marriages, widow re-marriage, and the Nautch question, we
also see him discuss customs and traditions, superstitions, use of English and the fascination towards it,
etc. Thus, making it culturally relevant and further justifying the use of Kanyasulkam in validating the
role of culture on humor. Persona identification is an important application of the schema proposed. For
instance, if character A has a tendency to sarcastically tease character B on most occasions, we gain an
insight into A’s sense of humor (SOH) as well the social function performed by A. While there have been
several studies that suggest that an SOH indicates positive personality traits such as self-actualising, self-
acceptance, and others (Maslow, 1954; Allport, 1961), the social function performed by A also provides
an understanding of A’s overall role in the story, therefore the character’s persona.

2 Related Work
Interest in the study of humor has faced steady growth since 1970 (McGhee P.E., 1989). This interest
in humor studies has led to a great deal of research on humor types and functions. In his paper on the
issues in conversational joking, Neal R. Norrick (2003) talks about the structure of humorous discourse,
the forms of conversational humor and its interpersonal functions, i.e., aggression vs. rapport. Two
of Marta Dynel’s studies, one based on a popular English sitcom, Friends (2011) and another on the
sitcom, House (2013), are deemed relevant to this study. While the former analyses cultural references,
the latter attempts to extract universal communicative phenomena that cause humor. Dirk Delabastita
(2005) presents in her work, an overview of the humorous scenes with bilingual and translation-based
situations from Shakespeare’s plays. Levisen (2014) uses Natural Semantic Metalanguage to compare
the Danish concept of ’sort humor’ (a highly culturally specific way of Danish communication) and the
English, ’black humor.’ To recognize humor and irony in tweets, Antonio Reyes et al. (2012) analyze
humor and irony to recognize these concepts in tweets. Agnese Augello et al. (2008) have worked on
building a chatbot that recognizes and generates humorous expressions. There have been continuous
efforts in the field of computer science for the comprehension (Binsted et al. 2006), detection (Taylor,
2009), production (Hempelmann et al., 2006), and recognition (Mihalcea et al., 2006) of conversational
humor.

3 Data and Annotation

The full text of the Telugu play, Kanyasulkam, is annotated by two people, A1 and A2. For the pre-
processing of the data, the whole text was split first by each character’s dialogue, and each utterance
by the character was further split into single sentences/segments. In the presence of poems, lists, etc.
the utterance remains as is, and this final output is used for annotation giving a total of 6645 segments
to be annotated. After developing the gold standard corpus (can be found here), 2710 utterances were
classified as humorous, 1782 were given the tag dialogue, 1881 conversational, and 892 benign. The
annotation was done with appropriate checkpoints after every 2000 segments to identify any new tech-
niques or revise the schema.

4 Annotation Schema
Types: Teasing(T), Retort(R), Banter(Ba), Schadenfreude(S)
Techniques: Dramatic Irony(DIrn), Sarcasm(Src), Satire(Str), Fallacious Reasoning(FR), Exaggera-
tion(Ex), Use of foreign language(FL), Allusion(A), Profanity(P), Other stylistic figures(O) (refer to
Table 5).
Note: A segment can be annotated with more than one technique.

35
 Figure 1: Hierarchical schema for Conversational humor

4.1 Level-1: Non-Conversational (NC)/ Conversational Humor (C)

This study uses Dynel’s (2009) conceptualization of conversational humor. As the dataset used here is a
play, it is of primal importance to note that the speaker’s intent may not be to cause humor. However, a
third-party present, or the metarecipients of the conversation, the audience/readers, may find it humorous.
It is a common phenomenon to make the audience laugh at the expense of a fictional character. This study
makes a distinction between conversational and non-conversational humor. In the latter, the humor does
not exist in the realm of verbality but rather in the domain of the situation (slapstick humor, a character’s
trait such as miserliness, stupidity, etc.) For example, when a cowardly character is badmouthing his
rival and the latter appears just then.

4.2 Level-2: Monologue (M)/ Dialogue (D)

The notion of ”dialogue” is taken in the Socratic dialogue sense: a conversation between two or more
people. In contrast, there are several definitions and types of monologues present: Dramatic monologue
(Finch, 2010), soliloquy (Shea, 1963), and inner monologue (Neuse, 1934). This study defines a mono-
logue as ”utterances by a single person/character in real-life/fictional with the assumption by the speaker
that there are no listeners present to hear their thoughts”. The distinction between humor found in a
dialogue and a monologue is made because it is recognized that if a person speaks to themselves with no
listeners present, it gives rise to certain types and techniques of humor compared to those occurring in a
dialogue. Take retorts, for instance. A retort is a sharp or witty remark in response to another’s utterance
(Sacks, 1992 [1972]). Hence, for this type of conversational humor, the first turn in an adjacency pair
must transpire (Sacks, 1992 [1972], 419).

4.3 Level-3: Benign (B)/ Non-Benign (NB)

The Benign Violation Theory examines the intersection at which the listener perceives a violation in a
joke as benign (McGraw, 2010). A joke is not a passive entity but is expressed by a person and perceived
by another. Hence, the relationship between the joke-teller, the joke, and the joke-listener must be
explored (Kant and Norman, 2019). In conversational humor, there is no notion of a ”joke” (”canned
joke” in Dynel (2009)). Humor is caused by the interlocutor’s spontaneous speech that may or may not
be humorous to the listeners present. For instance, in derision, reprimands, or put-downs, the speaker or
listener may not find it humorous; instead, a third-party present (or an audience) may find the utterance
humorous. In this study, it is important to note that the relationship between the author of the play and
the audience is not examined, but the latter is given the role of a passive listener.
karata: bAvA yIsammaMXaM ceswe nI koVMpaki aggeVtteswAnu.
agni: vIlYlammA SiKAwaraga, prawIgAdixakoVdukU wiMdipowullAga nAyiMtajeri nannanevAlYle

36
 Translation:
Karata: Brother-in-law, if you agree to this proposal, I will set your house on fire.
Agni: (An expletive directed at Karata’s mother), every son of a donkey, comes to my house to eat like
a glutton and ends up criticizing me.
Context:
Here, the interlocutors present in the scene do not find Agni’s utterance humorous as he only intends to
ridicule Karata and Venkamma. However, the metarecipients, the audience, are bound to find it amusing
(Dynel, 2009). This study augments the BVT by modifying the factors by which a joke can be labeled as
benign: (a) two contradictory norms of the relevant culture (b) a weak commitment to the violated norm,
or (c) the social distance between the interlocutors and the content of what is uttered (d) the intention of
the humor causer understood by the listener whether benign or not (Weiner 1993, 2009). By these four
conditions, the above example is labeled non-benign as it goes against the salient norm of respecting a
guest, and there exists no norm that states to insult a guest in the Telugu culture blatantly. Furthermore,
Agni’s intention is to solely deride his guest’s behavior.

4.4 Level-4: Types of conversational humor

4.4.1 Teasing (T)
In this study, teasing is considered to transpire when the speaker intends to be playful, to only nip at
the present listener non-aggressively. The main objective is to develop/strengthen the bond between the
speaker and the listener(s). Other than this benevolent intention, the speaker also uses an element of
”pretense” to tease (Clark, 1996; Dynel, 2009).
maXu: anyAyaM mAtalu AdakaMdi, Ayana yaMwa caxuvukunnAdu, Ayanaki yaMwapraKyAwi
vuMxi!
nedorepo goVppa vuxyogaM kAnEyyuMxi.
Translation:
Madhu: Don’t be unfair. He is a very learned man. He has a lot of fame as well. Very soon, he’ll land
himself a good job.
Context:
Madhu teases her client Ramappanthulu by praising another client of hers, Girisam. Rama detests
and is jealous of Girisam. Knowing this, Madhu seeks to elicit a reaction by exaggerating (a common
technique used in teasing) Girisam’s strength playfully.
4.4.2 Retort (R)
A retort takes place at the second turn in an adjacency pair where the purpose is to out-challenge or outwit
the other interlocutor(s) (Dynel, 2009) of the conversation by making a quick comeback (utilizing the
other’s behavior, personality, past, etc.) In this definition of retort, the speakers do not aim to collaborate
and engage in conversation jointly, rather undermine (non-benign) or challenge (in some cases, benign)
the listener (Holmes and Marra, 2002).
lubXA: mAmagAru hAsyAnikaMtunnArugAni, ninnoVxulwArA?
[. . . ]
rAma: alAgaddi peVttaMdi!
maXu: [. . . ]
gaddi gAdixalu wiMtAyi;
manuRyulu winaru.
Translation:
Lubdha: Uncle is just pulling your leg. Do you think he will forsake you?
[. . . ]
Rama: Teach her a lesson like that (Idiom with the literal meaning of ‘feed her grass’)
Madhu: [..]
Grass is eaten by donkeys, not people.

37
 Context:
Rama is reading a letter written by Girisam where the latter refers to him as a donkey. In response to
Rama’s suggestion of teaching Madhu a lesson, she mocks him indirectly by referring to the letter when
she says, ”Grass is eaten by donkeys, not by people”.
4.4.3 Banter (Ba)
If there is a continuous exchange of retorts and teasing in a multi-turn conversation, it is called banter
(Dynel, 2008; Norrick, 1993: 29). This rapid exchange of repartees is observed in interactions such as a
conversation between parents, coworkers at the office (example 2 in Holmes and Marra, 2002), etc. It is
important to note that Banter cannot be a hierarchical category encompassing Retort and Teasing as they
can also occur independently.
maXu: wAkattuvaswuva wappiMcuku pAripoweno?
kukkA, nakkA, kAxugaxA goVlusuluvesi kattadAniki?
karata: nI valallo paddaprANi mari wappiMcukupovadaM yalAga?
vAtiki vunna patuwvaM yevukku goVlusulakU vuMdaxu.
maXu: valalo muwyapu cippalupadiwe lABaMgAni, nawwagullalupadiwe mowacetu.
karata: yaMwasepU dabbu, dabbenA?
snehaM, valapU, anevi vuMtAyA?
maXu: snehaM mIlAtivAricota
Translation:
Madhu: What if you run away after pawning it?
You aren’t a dog or a fox to tie you with chains.
Karata: Can any living being be freed from your trap?
Its hold is stronger than that of any chain.
Madhu: Only if pearls are trapped, it is of any use. Getting a hold on rocks/shells will only increase
my burden.
Karata: Why are you always concerned about money? What about friendship, justice, etc.?
Madhu: Friendship with people like you (with sarcasm)
Context:
Madhu is hesitant to depart from her necklace, which Karata is asking for. Karata teases her by flattery
and hopes it will help in achieving his goal. However, Madhu retorts by indirectly comparing him to a
weed/stone. Subsequently, in response to his reprimand that she always talks about money, she retorts
using sarcasm once more.
4.4.4 Schadenfreude (S)
Schadenfreude is a German word that refers to the pleasure derived from another’s misfortune (Dijk et
al., 2009). It is the ”malicious joy” evoked by the downfall of others, mostly high achievers (Feather
and Sherman, 2002; Smith et al., 1996). This emotion is majorly associated with negative connotations,
(Smith et al., 2009, Leach et al., 2003; cf. Kuipers, 2014) All instances of Schadenfreude in this study
agree with the non-benign viewpoint. The intent of the utterance is to be truly abusive and denigrating
to the person it is directed towards (butt). An important aspect of annotation to note here is that, when a
segment is tagged as Schadenfreude, we need not laugh at the plight of the character whose utterance it
is.
PUta: AveVXavavuMte nAkeM kAvAli, vuMdakuMte nAkeM kAvAli.
vAdu nIkiccina yiravayi rUpAyalU yicceVy.
MaXu: yavadi kiccAvo vANNe adagavammA.
PUta: veVXavakanabadiwe sigapAyixIsi cIpurugattawo moVwwuxunu, yeVkkadaxAcAvevizti?
Translation:
Puta: I couldn’t care less about that idiot’s whereabouts. Just give me the 20 rupees that he gave you.
Madhu: Ask the person you gave it to.
Puta: I will cut his hair and thrash him with a broom if I find him, where did you hide him?

38
 Context:
Puta comes to Madhu’s house, searching for Girisam, who has run away with her money. On getting
no help from Madhu in finding his whereabouts, Puta is immensely angered, and humor is found in
Girisam’s plight.

4.5 Techniques:
4.5.1 Dramatic Irony (DIrn)
In a stage play setting, there exist two or more
levels of discourse, the author-audience/reader,
and the character-character level (Short and
Mick, 1996). When the character is portray-
ing pretense with one character, another charac-
ter may or may not be in on it, but the readers
necessarily are. Hence, other than the knowl-
edge that exists between the characters, the au-
Figure 2: Levels of discourse dience is also privy to knowledge only they pos-
sess (Kreuz and Roberts, 1993).
girI: AlrEt - gAni - nAkikkada cAlA vyavahAramulalo naRtaMvaswuMxe - munasabugAri pillalki
Salavullo pATAlceVpiweV PiptI rupij yiswAvaznnAru; ayinA nI viRayavEz yaMwa lAs vaccinA nenu ker
ceVyyanu.
Translation:
Giri: It’s alright. But I’ll incur a lot of losses here. The village head has promised to give 50 rupees
for tutoring his kids over the vacation. However, I do not care about any loss when it comes to you.
Context:
Girisam lies that he has been offered a valuable job opportunity but that he will reject it as he genuinely
cares for his pupil Venka. The audience knows that this is untrue as, before this conversation, Girisam
was plotting to take advantage of Venka’s economic resources.
As mentioned above, a requisite component of teasing is that of pretense. Nevertheless, teasing is not
mandatorily marked with dramatic irony as there is a difference in intent. In dramatic irony, the intent
is to dupe the listener by pretending to have values, attributes, etc. that the speaker does not possess.
Whereas, in teasing, the motivation is to reduce the social distance or to poke fun at the listener benignly.
4.5.2 Satire (Str)
Satire has been defined as the ridicule of a subject (a person, situation, or an institution) to point out its
faults (Beckson and Ganz, 1989). It does not need to be present at the scene of action. Studies on Telugu
literature have concluded satire to consist of 4 elements. Vyangyam, Chatuvulu, Prahasanams, and
Adhikshepam (Rao, 2004). The main features of each are sarcasm, ridiculing, intention of social reform,
and intention of teaching morals and ethics through severe criticism. The presence of these elements in
satire is only culturally significant, and such a clear-cut difference may not be found in English.
SiRyudu: ArneVllakomAtu poVswakaMpattukuMte koVwwaSlokAlu pAwaSlokAlu oVkkalA gkana-
padawAyi.
mAguruvugAriki xoVMdakAya kUra yiRTaM lexu, guruvugAri peVlYlAM peVratlo xoVMda-
pAxuMxani rojU AkUreV voVMduwuMxi.
bawikunnavAlYla yiRTaveVz yilA yeduswUMte caccinavAdi yiRTAyiRTAlwo yeMpani?
yIcaxuvikkadiwo cAliMci girISaMgAri xaggira nAlugiMgilIRu mukkalu nercukuMtAnu.
Translation:
Student: If I open my books once in 6 months, then the poems I have already learned, and the new
ones all look the same.
My teacher does not like ivy gourd curry. But his wife makes the same curry everyday owing to the
ivy gourd plant in their house.
If the likes of a person who is very much alive are not cared about, how do the likes of someone dead
matter?

39
 I should stop these lessons here and learn a few English words from Mr. Girisam.
Context:
Karata’s student is asked to learn a poem by heart where the poet talks about his likes and dislikes of
flowers and nature. The student is fed up by this mode of learning and feels it is pointless to learn about
a dead person’s likes and dislikes when his own guru’s likes are not cared about by the latter’s wife. This
example is a satire on the education being provided to the student by Karata. The element of satire being
used here is Chatuvu (ridicule).
4.5.3 Sarcasm (Src)
The difference between irony and sarcasm is fuzzy and is often misunderstood, given that they are in-
evitably bound to each other. However, the relationship between them remains unclear to native speakers
but is highlighted when a comparison is drawn between cultures or linguistic communities (Partington,
2006). Attardo, in his study, cites that sarcasm is an overly aggressive type of irony with more explicit
markers or cues and a clear target (Attardo, 2000). Studies also cite the difference between irony and sar-
casm as irony does not require the speaker’s intention and can be directed towards situations. However,
sarcasm must be deliberate and is a strictly verbal phenomenon (Haiman 1990,1998).
klArk: iMtiperU, sAkInU yemitaMdI?
BImA: emitayyA?
agni: Ayanaperu girISaM, maraMwakaMta nAkuweVliyaxu.
kaleV: cAbAR;
bAgAvuMxi!
avaXAnlugAri koVmArweVni yeVvado wIsukupoyinAdu.
kanaka vAdi vUrUperU yeVriginavAlYlu weVliyaceVyyavalasinaxani, xaMdorAkoVttiMci ge-
jatlo veyiMcaMdi.
Translation:
Clerk: What is his surname and address?
Bhima: What is it, man?
Agni: His name is Girisam. I do not know anything else.
Kale: Great. Sounds good. Let0 s get it published in the newspaper that someone kidnapped Mr. Agni0 s
daughter. And hence, anyone who knows his name and village should immediately inform us.
Context:
Agni goes to register a complaint against Girisam who runs away with the former’s daughter. When
Agni states that he knows only the first name and nothing else, the officer sarcastically praises him and
suggests that it would be great to publish this news in the Gazette and ask the public’s help to get to know
Girisam’s details.
4.5.4 Fallacious Reasoning (FR)
A fallacy is defined as an argument that has faulty reasoning (Gensler and Harry, 2010) either by inten-
tional pretense by the speaker or by genuine ignorance. In a conversational setting, a fallacy need not
be restricted to arguments presented by the speaker to reach a conclusion. However, the characteristics
of a conversation can be taken advantage of. For instance, the topic is diverted by speaking about an
unrelated topic, identifying a false cause and effect (Shewan and Edward, 1994), etc.
girISaM: veVrigud!
peVlYlanexi maMci paxArWavEzwe ”aXikasya aXikaM PalaM” annAdu ganaka cinnapillani
oVka musalAdiki peVlYlicesi, vAducaswe marodiki, maroducaswe marodiki, yilAga peVlYlimIxa peV-
lYli, peVlYlimIxa peVlYliayi, vIdixaggiro veVyyi, vAdixaggiro veVyyi, marodixaggira maroveVyyi, roVt-
teVmIxa neVyyi, newimIxa roVtteV lAga yekowravqxXigA kanyASulkaM lAgi, wuxaki nAlAMti buxXi-
vazMwuNnicUsi peVlYlAdiwe ceVppAv majA?
Translation:
Girisam: Very good. If marriage is a good thing, and since the more you do, the more you achieve, a
young girl should be married to an old man and once he dies, another man and if he dies, then another
one and so on while collecting a thousand from the first guy, then the next, then another, like butter on

40
bread and bread on butter, collect all the Kanyasulkam (bride price) and finally if she gets married to a
wise guy like me, isn’t that enjoyable?
Context:
Girisam pretends to agree that selling young girls for marriage is good for society when widow re-
marriage is allowed. He argues that for every man that dies naturally with old age over time if the child
is married and re-married to other older men, the father of the child gets money until the girl can marry
a sensible person like Girisam. It is evident that this is an example of ”non sequitur” fallacy, where the
premises are true, but the conclusion is false.
4.5.5 Utilizing a Foreign Language (FL)
Several studies have attempted to understand the motivations for using a foreign language to produce
humor (Siegel, 1995; Kim, 2006). Grosjean (1982) states that situations, messages, attitudes, and emo-
tions influence foreign language use. In Kanyasulkam, English is used sporadically only by one char-
acter, Girisam, to achieve his objective: to portray and distinguish himself among the characters as
well-educated.
agni: oVkkaxammidI yivvanu.
[. . . ]
ixaMwA topI vyavahAraMlA kanapaduwuMxi.
karata: [...]
girISaM: xisIj bArbaras, cUcAraMdI, jeVMtilmen anagA peVxxamaniRini yalA aMtunnAro!
Translation:
Agni: I will not spare even one penny
[...]
Karata: [...]
Girisam: This is barbarous. Did you see how he is talking to a gentleman, meaning, learned person!
Context:
Upon being accused of cheating by Agni, Girisam is angered. Here, knowing fully well that the
listeners do not understand English, Girisam still chooses to talk in English and then condescendingly
explains what he means. He does this to establish superiority over others as people who knew English in
those times were held in high regard.
4.5.6 Allusion (A)
(Norrick, 1989): Direct or indirect reference to an object or circumstance from a different context is
defined as an allusion.
karata: Ayanexo kurYrYavAdiwo yiMgilIRu mAtaMte puccakAyalaxoVMgaMte bujAlwaduvuzkunnattu
nImIxa peVttukuMtAveM?
Karata: If he is talking to his student in English, why are you getting involved like the watermelon
thief rubbing his shoulders (idiom) meaning, why are you letting yourself be caught red-handed by
getting angry and proving that you do not understand a word of it?
4.5.7 Profanity (P)
(Beers Fägersten, 2012): Profanity is defined as language that is considered as strongly impolite, rude,
or socially offensive.
girISaM: rAskeVl vulakalexu palakalexu sarekaxA moVhaM pakkaki wippi kaduppagiletattu
navvuwunnAdu.
Girisam: Not only did that rascal fail to support me during my lecture, he turned to his side and
laughed almost until his stomach burst.
4.5.8 Hyperbole/ Exaggeration (Ex)
(Norrick, 2004): It is the representation of an entity as more dramatic, better, or worse than it really is.
Hyperbole is a figure of speech using exaggeration.
girISaM: nene xAni hajbeVMdnEvuMte, nilabaddapAtuna nI waMdrini rivAlvarwo RUt ceSivuMxunu.
Girisam: If I were her husband, I would have shot your father with a revolver from where I stood.

41
 Although exaggeration necessarily has a pretense factor, any segment where exaggeration is identified,
”dramatic irony” is not marked.
4.5.9 Other identified techniques (O)
(Dynel, 2009): Such as simile, metaphor, etc. are also marked during annotation.
maXu: catlaki cAva nalupu, maniRiki cAva weVlupU. (simile)
Madhu: A person’s death is marked by white, like how a tree’s death is marked by black.

5 Disagreement Analysis
The validity of the tag set and their definitions are measured using Cohen’s Kappa (κ) (Cohen, 1960).
Although the annotators were asked to mark all levels of the hierarchical scheme, the inter-annotator
agreement (IAA) for level 2 (Monologue/Dialogue) was not measured as the definition for these cate-
gories provided no ambiguity. The annotation for level 1, Conversational vs. Non-Conversational Humor,
gave a Kappa value of 0.48 (moderate agreement). The disagreement emerged could be attested to the
variation in the perception of humor (Table 1). For instance, A2 could have found the character’s trait
(Non-conversational) humorous, whereas A1 identified a verbal technique in the speaker’s utterance,
causing disagreement. Annotation of level 3, Benign vs. Non-Benign Humor gave a Kappa value of 0.42
(moderate agreement). The disagreement exhibited can be due to the difference in perception of the be-
nignity of the utterance (Table 2). A1 could be aware of a salient norm that can be violated (Section 4.3),
whereas A2 is not producing disagreement.
Annotation of level 3 of the schema, types of Conversational Humor, resulted in a Kappa value of
0.49 (moderate agreement). Most of the disagreement shown in Table 2 (refer to Appendix), (Null,
<some type>) or (<some type>, Null), is due to the failure or success of labelling the type by one an-
notator or a difference in perception of humor itself. A significant overlap of types can only be observed
at (Retort, Teasing), which occurs at 11 segments. The dissimilarity of perception of the speaker’s intent
causes this overlap. Annotator A1 perceived that the speaker intends to outwit or challenge the listener,
whereas A2 perceived that the speaker only intends to pull the listener’s leg.
For the final level of the schema,
techniques of Conversational Humor
(Figure 3), it is worthy to note that each
segment can be marked with one or more
techniques, but it is not mandatory. If
level 1, 2, and 3 (Conversational/Non-
conversational, Monologue/Dialogue,
Benign/Non-Benign) are marked, then
level 4 (Type, Technique) is to be tagged
compulsorily. For this type of data, it
was considered best to use Krippen-
dorff’s Alpha (α) (Krippendorff, 2004)
for measuring the agreement between both
annotators, resulting in an alpha value of
Figure 3: Distribution of Techniques by A1 and A2 0.691. According to Krippendorff (2004),
tentative conclusions are acceptable where
alpha ≤ 0.667. The low agreement value can be attested to the following observations. The role of the
culture of the then period, and knowledge of the language itself to recognize allusion and wordplay
respectively contribute majorly. Failure to understand the invalidity of the argument presented by the
speaker leads to a Null tag in place of Fallacious Reasoning by the annotator. The common feature of
dramatic irony with both exaggeration (Section 4.5.6) and Fallacious Reasoning is the presenting of a
false statement to the listener, causing a grey area for annotation. Finally, the knowledge base possessed,
culture exposed to, the emotion experienced at the time of annotation influences the individual’s
subjectivity of humor (Attardo, 2003; Jiang et al., 2019; Martin and Ford, 2018), contributing to the
disagreement in the annotation of humor categories.

42
6 Conclusion and Future Work
The paper describes the work done on developing a fine-grained hierarchical annotation schema for
Conversational Humor. The annotation was performed on a relevant dataset, a prominent Telugu play
called Kanyasulkam. The inter-annotator disagreement highlighted the complexity of the task as well as
the domain itself. As mentioned in the introduction, the schema can be utilised for persona identification,
a use case especially beneficial for the literary field. For example, when analysing a Shakespearean
character, analysing their sense of humor may help the researcher recognize their pertinent traits. This
study also finds that inclusion of cultural nuances in the play has a significant effect on the perception of
humor. Further, this annotation schema can be applied to other culturally significant works by utilizing
the analysis provided in this work. However, when applying this schema to other works, it is to be
noted that the types and techniques listed here are non-exhaustive and more can be added based on the
language and cultural significance of the data being annotated. If a computer were to generate humor,
the traditional meaning would mean that it could generate ”jokes” (knock-knock jokes, etc.) but would
fail to generate conversational humor. It is believed that this work will aid in automating this process.

References
Richard Alexander. 1997. Aspects of verbal humour in English, volume 13. Gunter Narr Verlag.

Gordon W Allport. 1961. Pattern and growth in personality. New York: Holt, Rinehart and Winston.

Salvatore Attardo and Victor Raskin. 1991. Script theory revis (it) ed: Joke similarity and joke representation
model. Humor: International Journal of Humor Research.

Salvatore Attardo. 2000. Irony as relevant inappropriateness. Journal of pragmatics, 32(6):793–826.

Salvatore Attardo. 2003. Introduction: the pragmatics of humor. Journal of Pragmatics, 35(9):1287–1294.

Karl E Beckson and Arthur Ganz. 1989. Literary terms: A dictionary. Macmillan.

Roy R Behrens. 1977. Beyond caricature: On types of humor in art. Journal of Creative Behavior, 11(3):165–75.

Kim Binsted, Anton Nijholt, Oliviero Stock, Carlo Strapparava, G Ritchie, R Manurung, H Pain, Annalu Waller,
and D O’Mara. 2006. Computational humor. IEEE Intelligent Systems, 21(2):59–69.

Herbert H Clark. 1996. Using language. Cambridge university press.

Jacob Cohen. 1960. A coefficient of agreement for nominal scales. Educational and psychological measurement,
20(1):37–46.

Dirk Delabastita. 2005. Cross-language comedy in shakespeare. Humor, 18(2):161–184.

Marta Dynel. 2008. No aggression, only teasing: The pragmatics of teasing and banter. Lodz papers in pragmat-
ics, 4(2):241–261.

Marta Dynel. 2009. Beyond a joke: Types of conversational humour. Language and Linguistics Compass,
3(5):1284–1299.

Marta Dynel. 2011. I’ll be there for you: On participation-based sitcom humour. The pragmatics of humour
across discourse domains, 311:333.

Marta Dynel. 2013. Humorous phenomena in dramatic discourse. The European Journal of Humour Research,
1(1):22–60.

Kristy Beers Fägersten. 2012. Who’s swearing now? The social aspects of conversational swearing. Cambridge
Scholars Publishing.

Norman T Feather and Rebecca Sherman. 2002. Envy, resentment, schadenfreude, and sympathy: Reactions
to deserved and undeserved achievement and subsequent failure. Personality and Social Psychology Bulletin,
28(7):953–961.

Matthew Finch. 2010. Thrown voices: A series of dramatic monologues, with a discussion of the genre.

43
Harry J Gensler. 2010. The A to Z of Logic. Number 169. Rowman & Littlefield.

François Grosjean. 1982. Life with two languages: An introduction to bilingualism. Harvard University Press.

Christian Hempelmann, Victor Raskin, and Katrina E Triezenberg. 2006. Computer, tell me a joke... but please
make it funny: Computational humor with ontological semantics. In FLAIRS Conference, volume 13, pages
746–751.

Janet Holmes and Meredith Marra. 2002. Having a laugh at work: How humour contributes to workplace culture.
Journal of pragmatics, 34(12):1683–1710.

Tonglin Jiang, Hao Li, and Yubo Hou. 2019. Cultural differences in humor perception, usage, and implications.
Frontiers in psychology, 10:123.

Leo Kant and Elisabeth Norman. 2019. You must be joking! benign violations, power asymmetry, and humor in a
broader social context. Frontiers in psychology, 10:1380.

Eunhee Kim. 2006. Reasons and motivations for code-mixing and code-switching. Issues in EFL, 4(1):43–61.

Roger J Kreuz and Richard M Roberts. 1993. On satire and parody: The importance of being ironic. Metaphor
and Symbol, 8(2):97–109.

Klaus Krippendorff. 2018. Content analysis: An introduction to its methodology. Sage publications.

Giselinde Kuipers et al. 2014. Schadenfreude and social life: a comparative perspective on the expression and
regulation of mirth at the expense of others. Schadenfreude. Understanding Pleasure at the Misfortune of
Others. Cambridge University Press: Cambridge, pages 259–294.

Colin Wayne Leach, Russell Spears, Nyla R Branscombe, and Bertjan Doosje. 2003. Malicious pleasure:
Schadenfreude at the suffering of another group. Journal of personality and social psychology, 84(5):932.

Carsten Levisen. 2014. The story of “danish happiness”: Global discourse and local semantics. International
Journal of Language and Culture, 1(2):174–193.

Rod A Martin and Thomas Ford. 2018. The psychology of humor: An integrative approach. Academic press.

Abraham H Maslow. 1954. The instinctoid nature of basic needs. Journal of personality, 22(3):326–347.

Paul E McGhee. 1989. Introduction: Recent developments in humor research. Journal of children in contempo-
rary society, 20(1-2):1–12.

A Peter McGraw and Caleb Warren. 2010. Benign violations: Making immoral behavior funny. Psychological
science, 21(8):1141–1149.

John C Meyer. 2000. Humor as a double-edged sword: Four functions of humor in communication. Communica-
tion theory, 10(3):310–331.

Rada Mihalcea and Carlo Strapparava. 2006. Learning to laugh (automatically): Computational models for humor
recognition. Computational Intelligence, 22(2):126–142.

Kerry Mullan and Christine Béal. 2018. Introduction: Conversational humor: Forms, functions and practices
across cultures. Intercultural Pragmatics, 15(4):451–456.

Werner Neuse. 1934. ” erlebte rede” und” innerer monolog” in den erzählenden schriften arthur schnitzlers.
Publications of the Modern Language Association of America, pages 327–355.

Neal R Norrick. 1989. Intertextuality in humor. Humor, 2(2):117–140.

Neal R Norrick. 1993. Conversational joking: Humor in everyday talk. Indiana University Press.

Neal R Norrick. 2003. Issues in conversational joking. Journal of pragmatics, 35(9):1333–1359.

Neal R Norrick. 2004. Hyperbole, extreme case formulation. Journal of Pragmatics, 36(9):1727–1739.

Alan Partington. 2006. The linguistics of laughter: A corpus-assisted study of laughter-talk. Routledge.

Durga Srinivasa T Rao. 2004. Problems of translatiing satire from english to telugu and vice versa: an evaluation.

44
Victor Raskin. 1979. Semantic mechanisms of humor. In Annual Meeting of the Berkeley Linguistics Society,
volume 5, pages 325–335.
Victor Raskin. 1985. Semantic theory of humor. In Semantic Mechanisms of Humor, pages 99–147. Springer.

Antonio Reyes, Paolo Rosso, and Davide Buscaldi. 2012. From humor recognition to irony detection: The
figurative language of social media. Data & Knowledge Engineering, 74:1–12.
Harvey Sacks and Gail Jefferson. 1995. Lectures on conversation.
James M Shea. 1963. Convention and invention: Soliloquy in Shakespearean tragedy. University of Windsor.

Edward Shewan. 1994. Applications of Grammar: Principles of Effective Communication, Book 4. Christian
Liberty Press.

Mick Short. 2018. Exploring the language of poems, plays and prose. Routledge.
Jeff Siegel. 1995. How to get a laugh in fijian: Code-switching and humor. Language in Society, pages 95–110.
Richard H Smith, Terence J Turner, Ron Garonzik, Colin W Leach, Vanessa Urch-Druskat, and Christine M
Weston. 1996. Envy and schadenfreude. Personality and Social Psychology Bulletin, 22(2):158–168.
Richard H Smith, Caitlin AJ Powell, David JY Combs, and David Ryan Schurtz. 2009. Exploring the when and
why of schadenfreude. Social and Personality Psychology Compass, 3(4):530–546.

Julia M Taylor. 2009. Computational detection of humor: A dream or a nightmare? the ontological semantics
approach. In 2009 IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intelligent Agent
Technology, volume 3, pages 429–432. IEEE.

Joanna Thornborrow and Shân Wareing. 1998. Patterns in language: An introduction to language and literary
style. Psychology Press.
Wilco W Van Dijk, Jaap W Ouwerkerk, and Sjoerd Goslinga. 2009. The impact of deservingness on schadenfreude
and sympathy: Further evidence. The Journal of Social Psychology, 149(3):390–392.
Bernard Weiner. 1993. On sin versus sickness: A theory of perceived responsibility and social motivation. Amer-
ican psychologist, 48(9):957.
Bernard Weiner. 2006. Social motivation, justice, and the moral emotions: An attributional approach. Psychology
Press.

45
 Appendix

A2
Verbal Situational NULL Total
Verbal 1559 120 566 2425
A1
Situational 191 354 284 829
Null 290 208 2893 3391
Total 2040 682 3923 6645

Table 1: Cohen’s Kappa for Level-1

A2
Benign Non-Benign Null Total
Benign 336 9 127 472
A1
Non-benign 412 802 739 1953
Null 220 261 3739 4220
Total 968 1072 4605 6645

Table 2: Cohen’s Kappa for Level-3

A2
Teasing Retort Banter Schadenfreude Null Total
Teasing 62 1 0 0 38 101
Retort 11 10 0 0 13 33
A1
Banter 1 0 131 0 6 138
Schadenfreude 3 5 0 93 185 286
Null 123 10 12 107 5835 6087
Total 200 26 143 200 6076 6645

Table 3: Cohen’s Kappa for Level-4

Level-1 Level-2 Level-3 Level-4

Conversational Monologue Benign Types Techniques
Non-Conversational Dialogue Non-Benign Teasing Dramatic Irony
Banter Sarcasm
Retort Satire
Schadenfreude Fallacious Reasoning
Exaggeration
Use of Foreign Language
Allusion
Profanity
Other Stylistic Figures

Table 4: Hierarchical Annotation Schema

46
Tag Description Example
C Conversational Conversation has one speaker and two listeners.
NC Non-Conversational The Three Stooges getting poked in the eye or thrown
pies at their faces.
M Monologue Only one speaker present and no listeners.
D Dialogue Conversation has one speaker and three listeners.
B Benign [A short person can’t reach a shelf by a wide mar-
gin] A friend says, “If only you were an inch taller.”
NB Non-Benign “The woman who is yelling in the street is a rascal
that bites men”
T Teasing [A woman spills her drink] Her boyfriend says, “Let
me grab a sippy cup for you”
R Retort “I’m sorry but I don’t speak bullshit.”
Ba Banter A series of teases and retorts between speakers.
S Schadenfreude ”Somebody stole my lunch out of the fridge at work
today. The worst part about it... I’m working from
home.”
DIrn Dramatic Irony [A character is known to be promiscuous] He says,
“None can be loyal to a woman as I am”
Sar Sarcasm [Torrential rain on an expected sunny day] “Oh
what warm weather!”
Str Satire “People say jokes are dead. But one can be found
alive and kicking in the White House.”
FR Fallacious Reasoning “I never generalize because everyone who does is a
hypocrite.”
Ex Exaggeration “How are you still hungry? You have a bottomless
pit for a stomach.”
FL Use of Foreign language [A mother asks her son to come home] He replies,
“Je ne comprends pas!”
A Allusion “Don’t act like Romeo in front of her!”
P Profanity “The idiotic excuse of a brother I have has no sense
of decency!”
O Other identified techniques “She was as tall as a six-foot-two-inch tree.”

Table 5: Humor Tagset

47