=Paper=
{{Paper
|id=Vol-2481/paper73
|storemode=property
|title=Annotation and Analysis of the PoliModal Corpus of Political Interviews
|pdfUrl=https://ceur-ws.org/Vol-2481/paper73.pdf
|volume=Vol-2481
|authors=Daniela Trotta,Sara Tonelli,Alessio Palmero Aprosio,Annibale Elia
|dblpUrl=https://dblp.org/rec/conf/clic-it/TrottaTAE19
}}
==Annotation and Analysis of the PoliModal Corpus of Political Interviews==
https://ceur-ws.org/Vol-2481/paper73.pdf
Annotation and Analysis of the PoliModal Corpus
of Political Interviews
Daniela Trotta Sara Tonelli Alessio Palmero Aprosio Annibale Elia
Università di Salerno FBK FBK Università di Salerno
dtrotta@unisa.it satonelli@fbk.eu aprosio@fbk.eu elia@unisa.it
Abstract interviews in Italian, and analyze how the combi-
nation of verbal and non-verbal elements can shed
English. In this paper, we present the first new light into political agendas and politicians’ at-
available corpus of Italian political inter- titude. By ‘multimodal’ we mean that the corpus is
views with multimodal annotation, con- composed of manual transcriptions of interviews
sisting of 56 face-to-face interviews taken broadcast on TV and annotated with information
from a political talk show. We detail the not only about the linguistic structure of the utter-
annotation scheme and we present a num- ances but also about non-verbal expressions2 .
ber of statistical analyses to understand the The corpus, which we call PoliModal, addresses
relation between these multimodal traits the need to make up for the lack of Italian lin-
and language complexity. We also ex- guistic resources for political-institutional com-
ploit the corpus to test the validity of exist- munication and is annotated in XML following
ing studies on political orientation and lan- the standard for the transcriptions of speech TEI
guage use, showing that results on our data Guidelines for Electronic Text Encoding and In-
are not as clear-cut as on English ones.1 terchange3 . In all transcripts, interviewers, in-
terviewees and other guests’ turns have been en-
1 Introduction riched with the manual annotation of non-lexical
In the context of a political interview, the host, and semi-lexical aspects such as breaks, inter-
typically a journalist, acts as a representative of ruptions, false starts, overlaps, interjections, etc.
the audience. This means that, if a politician man- Furthermore, additional linguistic traits related to
ages to convince or deal with the criticism that the language complexity, use of pronouns and per-
host addresses, then her/his trustworthiness, reli- sons’ mentions have been automatically tagged,
ability and credibility will be easily established. enabling an in-depth analysis of speakers’ atti-
In this situation, a politician is judged not only tude and communication strategy. In this work
based on one’s arguments and rhetorical choices, we present not only the corpus, which is made
but also on the attitude, self-confidence, and in freely available at the link https://github.
general on an overall convincing behaviour. For com/dhfbk/InMezzoraDataset, but also
example, if a politician seems to be conversation- an analysis that, combining verbal and non-verbal
ally dominant and manages interruptions to a sat- elements, shows how these traits contribute to
isfactory degree, it is more likely that the host, making an interview more or less convincing.
and therefore the audience, will be convinced by
2 Related work
the arguments put forward by the interviewee. For
this reason, analysing the combination of verbal In recent years, political language has received in-
and non-verbal elements in a political interview creasing attention, especially in the Anglo-Saxon
could be very interesting for scholars in political
2
science and communication science, and in gen- According to (Allwood, 2008): “The basic reason for
collecting multimodal corpora is that they provide material
eral to study consensus mechanisms. In this light, for more complete studies of ‘interactive face-to-face shar-
we present the first multimodal corpus of political ing and construction of meaning and understanding’ which is
what language and communication are all about”.
1 3
Copyright c 2019 for this paper by its authors. Use per- P5: Guidelines for Electronic Text Encoding and In-
mitted under Creative Commons License Attribution 4.0 In- terchange. See more https://tei-c.org/release/
ternational (CC BY 4.0). doc/tei-p5-doc/en/html/TS.html#TSSAPA
and American world, where it is possible to have persuasion and interruption during political inter-
free access to speech transcriptions from govern- views. Their work, however, is mainly aimed
ment portals and personal foundation websites, at studying the strategies for conversational dom-
e.g. White House portal, William J. Clinton Foun- inance, and annotate specific traits accordingly.
dation, Margaret Thatcher Foundation. This has Our work, instead, is more general, includes a dif-
fostered research on political and media commu- ferent set of tags and integrates also automatic lin-
nication and persuasion strategies (Guerini et al., guistic features.
2010; Esposito et al., 2015).
However, not all languages are well represented 3 Description of the PoliModal corpus
in this kind of studies. According to LRE Map4 The PoliModal corpus includes the transcripts of
there are currently 24 monolingual corpora for 56 TV face-to-face interviews of 14 hours - taken
Italian, two of which concern spoken language, from the Italian political talk show “In mezz’ora in
i.e. VoLIP (Alfano et al., 2014) and LUNA cor- più” broadcast from 24 September 2017 to 14 Jan-
pus (Dinarelli et al., 2009), and one multimodal, uary 2018.The show follows a fixed format, with
named ImagAct-ItalWorNet-Mapping (Bartolini interviews conducted by a journalist, Lucia An-
et al., 2014); no entry includes an Italian corpus nunziata, to a guest, typically a prominent figure in
for the political domain. Furthermore, researchers the political or cultural scene. A secondary guest
in Italian politics have mainly focused on political may participate as well, usually a second journal-
communication in the verbal modality, evaluating ist to comment on the debate. Each interview is
monological discourse (Bolasco et al., 2006; Ce- done in the same limited time frame, 30 minutes,
droni, 2010; Longobardi, 2010; Catellani et al., and no audience is present, so that applause and
2010; Bongelli et al., 2010; Zurloni and Anolli, any other type of reactions are not included in the
2010; Sprugnoli et al., 2016; Moretti et al., 2016) corpus.
to study a politician’s lexical, textual or rhetori- The audio signal has been transcribed us-
cal patterns. An exception is the work by Salvati ing a semi-supervised speech-to-text methodology
and Pettorino (2010), that diachronically analy- (Google API + manual correction). All hesita-
ses some of the suprasegmental aspects of Berlus- tions, repetitions and interruptions of the original
coni’s speeches from 1994 to 2010. The corpus, interview have been included. The output has been
however, is not available for further studies. further segmented into turns, and punctuation has
Concerning political corpora developed specif- been added, mainly to delimit sentence boundaries
ically for conversation analysis, Bigi et al. (2011) when they were not ambiguous.
present a multimodal corpus of political debates at It is important to note that, even if transcription
the French National Assembly, on May 4th, 2010 seems to be an objective task, it involves a cer-
and introduce an annotation scheme for a politi- tain degree of interpretation. Indeed, the inclusion
cal debate dataset which is mainly in the form of of the punctuation necessary to make the writing
video and audio annotations. Navarretta and Pag- comprehensible, as well as the selection of non-
gio (2010) deal with the identification of interlocu- verbal messages and non-verbal expressions (in-
tors via speech and gestures in annotated televised terjections, laughter, unfinished words, etc.) are
political debates in British and American English. interpretative choices aimed at revealing a sense.5
Other papers have focused primarily on visual as- Therefore, in the case of ambiguous sentences,
pects (gaze, gestures, facial expressions) of com- they have been identified manually, mainly look-
municative interaction during political talk shows ing at the context of the enunciation. According to
or parliamentary speeches (D’Errico et al., 2010). (Ducrot, 1995), in fact, it is not possible to under-
The most similar approach to ours is presented stand a communicative act without knowing the
in Koutsombogera and Papageorgiou (2010). The context in which it occurs. The context is there-
authors analyse a Greek multimodal corpus of 10 fore essential to choose one of the possible inter-
face-to-face television interviews focusing on non- pretations of ambiguous expressions.
verbal aspects in order to study the attempts of
5
As (Portelli, 1985) reminds us: “La punteggiatura serve
4
LRE Map is a mechanism intended to monitor the use sia a scandire il ritmo che a gerarchizzare sintatticamente
and creation of language resources by collecting information il discorso; non sempre le due funzioni coincidono, per cui
on both existing and newly-created resources, free available trascrivendo si è costretti spesso optare per l’una a danno
at http://lremap.elra.info/ dell’altra”
In PoliModal, annotation has been done using corresponding tag is del. We include these in
XML as markup language and following the TEI our annotation since several past studies (Simone,
standard for Speech Transcripts in terms of ut- 1990; Bazzanella, 1992; Tannen, 1989) high-
terances. The linguistic resource has currently lighted their importance in spontaneous speech,
100,870 tokens and includes interviews to politi- mentioning in particular the role of repetitions
cians covering all the Italian political spectrum in controlling the in-progress textual design of
(from the extreme right movement Casa Pound to speech (Voghera, 2001).
the liberal and progressive Partito Radicale). Be- (e) overlap: this phenomenon is present when
side politicians, also a small number of people the speaker conveys (in a verbal or non-verbal
with different backgrounds (students, academics, manner) that he/she is about to finish his/her turn
judges, economists, etc.) has been interviewed and and the co-locutor starts speaking so that there
is therefore included in the corpus. is a slight overlap of utterances. Overlaps can
For each interview the following information be competitive, when the overlapper disrupts the
was manually annotated and is included in the speech and can be perceived as intrusive by domi-
XML resource file: nating the conversation, and cooperative, when the
(a) metadata: these include useful information goal of the overlapper is to maintain the flow of
for a quick identification of transcriptions, for ex- the turns and add to the conversation with further
ample the tools used for the transcription, a link comments (Truong, 2013).
to the interview, the owner account, the title of the
talk show, the date of airing, the guests, etc. 4 Corpus Analysis
(b) pause: this tag is used to mark a pause ei- In this section, we analyse several linguistic di-
ther between or within utterances. Speakers differ mensions that can be either automatically ex-
very much in their rhythm and in particular in the tracted or derived from the corpus annotation, and
amount of time they leave between words, so the that can contribute to better understand typical
following element is provided to mark occasions traits of political communication.
where the transcriber judges that a speech has been
paused, irrespective of the actual amount of si- 4.1 Statistics of Non-Verbal Traits
lence. Several studies have converged on the con- We first group the politicians in our corpus into
clusion that we alternate between planning speech political parties, and then analyse those that are
and implementing our plans. Indeed, as shown in represented by least 3 politicians: Forza Italia, a
(Henderson et al., 1966), participants to interviews conservative center-right political party (3 inter-
typically show a cycle of hesitation and fluency, al- views), Lega Nord, a right-wing political party
though the ratio of speech to silence varies among often targeting immigrants (5 interviews), Movi-
speakers. mento 5 Stelle, a populist citizens’ movement (3
(c) vocal: with this tag we mark any vocal- interviews) and Partito Democratico, a moder-
ized but not necessarily lexical phenomenon, for ate centre-left political party (9 interviews). An
example non-lexical expressions (i.e. burp, click, overview of the distribution of non-verbal traits in
throat, etc.) and semi-lexical expressions (i.e. ah, the PoliModal corpus for each party is reported
aha, aw, eh, ehm etc.). These traits have been asso- in Fig. 1. Although the graph shows some dif-
ciated with the fact that linguistic planning is very ferences in the frequency of occurrences, they are
cognitively demanding, and it is difficult to plan an not statistically significant, also because of the rel-
entire utterance at once (Lindsley, 1975). There- atively small number of interviews considered in
fore, hesitation pauses and similar vocal phenom- the study. Also, the standard deviation for the av-
ena may be useful to perform a careful lexical re- erages tends to be high, showing high differences
trieval, since past studies (Levelt, 1983) found that among interviewees of the same party. For ex-
pauses occurred more often before low-frequency ample, politicians of Lega Nord make on aver-
words than before high frequency ones. age more pauses, but the range goes from 0.286
(d) del: this tag covers different phenomena of per turn (Roberto Maroni) to 0 (Luca Zaia). Sim-
speech management, specifically false starts, repe- ilarly, non-lexical and semi-lexical expressions,
titions and truncated words. Since they are marked marked as vocal, are on average more frequent for
in the TEI Guidelines as ‘editorially deleted’, the PD politicians, but range from 1.25 per turn (En-
rico Letta) to 0.10 (Matteo Renzi). These results that in our case the hypothesis by Schoonvelde et
show that differences pertain more to single per- al. (2019) is not confirmed, with the three highest
sons and conversational style than to political ori- ttr values belonging to politicians from three dif-
entation. An exception is given by overlaps, for ferent parties: Forza Italia (Mariastella Gelmini,
which the three politicians of M5Stelle (Alessan- 0.87 ttr), Lega Nord (Matteo Salvini, 0.82) and PD
dro Di Battista, Luigi Di Maio, Giancarlo Cancel- (Michele Emiliano, 0.82).
leri) all show a frequency above average, suggest-
ing that it may be connected with the communica-
tion strategy of the members of Movimento.
Figure 2: Avg. ttr and conceptual density per po-
litical party
Figure 1: Distribution of traits per political party A second hypothesis we want to test is the one
(avg. number of occurrences per turn). introduced in the work by Cichocka et al. (2016),
where the authors show that Republican presidents
used a higher proportion of nouns than Demo-
4.2 Political orientation and Language Use
cratic presidents, while there were no reliable dif-
A second analysis we carry out is related to exist- ferences in the use of verbs or adjectives. The au-
ing works about the use of linguistic features re- thors suggest that, compared to liberals, conserva-
lated to political orientation. In particular, a re- tive politicians are more inclined to use parts of
cent study by Schoonvelde et al. (2019) has anal- speech that stress clarity and predictability (such
ysed more than 380,000 speeches from five dif- as nouns) and reduce uncertainty and ambiguity
ferent Parliaments, and has proven that ideologi- (such as verbs or adjectives). We therefore com-
cally conservative politicians use a less complex pute the average number of nouns, adjectives and
language than liberal ones (this result is however verbs per political party and compare them. Sim-
less clear for economic left-right ideology). Since ilar to the previous analysis, averages are all in
these findings were not tested on Italian political the same range and there is no statistically signif-
documents, we carry out a comparison using the icant difference among parties. However, some
collected transcripts. In order to analyse the com- of the results are in line with Cichocka et al.’s
plexity of the language used by each politician we study, with PD showing a slightly lower number
computed the type-token ratio and the average lex- of nouns on average (and Valeria Fedeli being the
ical density, i.e. the number of content words di- politician with the lowest noun ratio, 0.16). Also,
vided by the total number of tokens. We do not Matteo Salvini and Luigi di Maio are the politi-
take into account the Gulpease index (Lucisano cians with the highest use of nouns, 0.22 per to-
and Piemontese, 1988), which is the de-facto stan- ken on average. A further evidence in favour of
dard metric of readability in Italian, because it was these results are the statistics obtained on the use
meant for written documents and heavily relies of content words, in particular on the percentage
on sentence length, a boundary that is not always of nouns, verbs, adverbs and adjectives, reported
present in transcripts. in Fig 3. We consider the five politicians with
Fig.2 shows the average type-token ratio and the highest number of turns in the corpus (see
conceptual density per political party. There are Table 1): Alessandro Di Battista (Movimento 5
almost no variations among the parties, with small Stelle), Carlo Calenda (PD), Matteo Renzi (PD),
standard deviations. This comparison suggests Angelino Alfano (Popolo delle Libertà), Matteo
Salvini (Lega). The figure confirms that Matteo
Salvini is the politician using the most nouns on
average, in line with the findings by Cichocka et
al. (2016). Carlo Calenda, instead, is the politi-
cian that on average uses most verbs and adverbs,
conveying more uncertainty and ambiguity than
all the other politicians including Matteo Renzi.
Figure 4: Avg. nouns per political party
that lexical density shows a moderate negative cor-
relation with repetitions (n=13, r=–0.51), trunca-
tions (r=–0.46) and non-lexical and semi-lexical
expressions (r=–0.43). On the contrary, it has
a moderate positive correlation with the average
Figure 3: Use of nouns, verbs, adjectives and number of pauses (r=0.49). This result suggests
adverbs for each politician (% over all content that, among the manual traits, pauses are used as
words) a linguistic device and are an indicator of a good
control of the conversation. Therefore, they are
The fact that the two studies considered do not more often used by politicians showing a high lex-
find a clear confirmation in our corpus, where the ical density, i.e. the ability to convey concepts in a
differences among the parties are rather blurred, concise way, which is crucial especially during TV
may have three possible explanations: i) this cor- interviews. The other manually annotated traits,
pus may be too small to test the above hypothe- instead, seem to be more frequent in speeches that
ses. Its expansion is indeed already in progress; are less organised, for which the management of
ii) the hypotheses do not actually hold in our case, the discourse is less efficient.
i.e. in the Italian political scene it is not true that Among the politicians considered in this study,
liberals use more complex language and tend to Carlo Calenda makes on average the highest num-
use less nouns than conservatives; or iii) the four ber of pauses (0.27 per turn on average, with
parties considered cannot be straightforwardly di- a lexical density of 0.579), followed by Giulio
vided into liberals and conservatives, and there are Tremonti (0.16 pauses per turn, 0.585 lexical den-
different positions inside the same party. sity).
4.3 Relation between verbal and non-verbal
5 Conclusions
traits
A third analysis is aimed at studying the correla- In this work, we present PoliModal, the first
tion between non-verbal traits and language com- freely-available multimodal corpus of political in-
plexity. We therefore focus on the interviews that terviews, manually annotated with six non-verbal
have a minimal length of 50 turns. The list of traits. The corpus covers 56 interviews, where
politicians and corresponding count of annotated each guest is associated with a role (for non politi-
traits is reported in Table 1. Again, for complexity cians) or a political party. We also present a first
we consider type-token ratio and conceptual den- statistical analysis of the traits and their associa-
sity. tion with language complexity and with the speak-
We perform an analysis of the correlation be- ers’ political orientation.
tween language complexity and the six non-verbal In the future, we plan to start from the anno-
traits manually annotated in the interviews, nor- tated material not only to extend the corpus, but
malised by the number of turns uttered by each also to investigate other aspects of political com-
politician. While type-token ratio (TTR) does not munication. For example, the choice to note non-
correlate with any of the manual traits, we found verbal expressions is motivated by the will to study
Guest Turn Repetition FalseStart Truncation Overlap Pause Non-lexical Semi-lexical
Alessandro Di Battista 203 24 14 34 76 19 9 66
Carlo Calenda 137 10 13 1 48 37 1 34
Matteo Renzi 187 40 19 69 25 0 3 16
Walter Veltroni 55 16 12 10 11 0 2 8
Simone Di Stefano 91 20 5 15 23 0 0 4
Pierluigi Bersani 92 30 0 20 15 1 14 24
Angelino Alfano 100 17 3 3 31 9 2 22
Giulio Tremonti 56 8 0 0 14 9 2 6
Matteo Orfini 67 10 0 0 21 1 2 8
Luigi Di Maio 74 14 0 14 32 0 4 11
Matteo Salvini 1 57 13 0 11 19 3 2 14
Matteo Salvini 2 86 19 3 3 30 13 7 19
Pier Carlo Padoan 67 5 1 7 13 8 13 21
Table 1: Corpus statistics related to the 13 interviews included in our study
the strategies of persuasion used by the speak- the Ninth International Conference on Language Re-
ers. According to Poggi (2005), persuasion strate- sources and Evaluation (LREC’14), Reykjavik, Ice-
land, may. European Language Resources Associa-
gies are multimodal constructs because politicians
tion (ELRA).
– specifically in televised political interviews –
attempt to persuade their supporters not only by Jens Allwood. 2008. Multimodal Corpora. In
their discursive style and argumentative speech, Lüdeling, A. & Kytö, and M., editors, Corpus Lin-
guistics. An International Handbook, pages 207–
but also through their personality and their inter- 225. Mouton de Gruyter.
actional behaviour. In the context of a political
interview, persuasion is related to conversational Roberto Bartolini, Valeria Quochi, Irene De Felice,
dominance, i.e. a speaker’s tendency to control Irene Russo, and Monica Monachini. 2014. From
synsets to videos: Enriching italwordnet multi-
the other speaker’s conversational actions over the modally. In Nicoletta Calzolari (Conference Chair),
course of an interaction (Itakura, 2001), which is Khalid Choukri, Thierry Declerck, Hrafn Lofts-
made evident through the kind of non-verbal ex- son, Bente Maegaard, Joseph Mariani, Asuncion
pressions annotated in our corpus. Moreno, Jan Odijk, and Stelios Piperidis, editors,
Proceedings of the Ninth International Conference
Finally, since at the moment only one annotator on Language Resources and Evaluation (LREC’14),
has performed the transcription, segmentation and Reykjavik, Iceland, may. European Language Re-
tagging task, we plan to compute inter-annotator sources Association (ELRA).
agreement in the near future. The annotation task Carla Bazzanella. 1992. Aspetti pragmatici della
addressed so far falls – from a qualitative point ripetizione dialogica. In Giovanni Gobber, editor,
of view – in the first of the general types identi- Linguistica pragmatica, volume XXIV of Atti SLI,
fied by (Mathet et al., 2015), in which the subjec- pages 433–454, Roma. Bulzoni.
tive interpretation is limited. Indeed, it deals with Brigitte Bigi, Cristel Portès, Agnès Steuckardt, and
the “identification of units” (Krippendorff, 2018), Marion Tellier. 2011. Multimodal annotations and
in which the annotator, given a written or spoken categorization for political debates. In ICMI Work-
text, must identify the position and boundary of shop on Multimodal Corpora for Machine learning,
pages 1–4.
linguistic elements (e.g. identification of prosodic
or gestural units, topic segmentation). We there- Sergio Bolasco, Nora Galli de’Paratesi, and Luca Giu-
fore expect agreement to be at least fair, but we liano. 2006. Parole in libertà: un’analisi statis-
plan to measure it using standard metrics. tica e linguistica dei discorsi di Berlusconi. Mani-
festolibri.
Ramona Bongelli, Ilaria Riccioni, and Andrzej
References Zuczkowski. 2010. Certain-uncertain, true-false,
good-evil in italian political speeches. In Interna-
Iolanda Alfano, Francesco Cutugno, Aurelio De Rosa, tional Workshop on Political Speech, pages 164–
Claudio Iacobini, Renata Savy, and Miriam Voghera. 180. Springer.
2014. Volip: a corpus of spoken italian and a
virtuous example of reuse of linguistic resources. Patrizia Catellani, Mauro Bertolotti, and Venusia Cov-
In Nicoletta Calzolari (Conference Chair), Khalid elli. 2010. Counterfactual communication in pol-
Choukri, Thierry Declerck, Hrafn Loftsson, Bente itics: Features and effects on voters. In Interna-
Maegaard, Joseph Mariani, Asuncion Moreno, Jan tional Workshop on Political Speech, pages 75–85.
Odijk, and Stelios Piperidis, editors, Proceedings of Springer.
Lorella Cedroni. 2010. Politolinguistics. Towards a Pietro Lucisano and Maria Emanuela Piemontese.
New Analysis of Political Discourse. In Interna- 1988. Gulpease. Una formula per la predizione della
tional Workshop on Political Speech, pages 220– difficoltà dei testi in lingua italiana. Scuola e Città,
232. Springer. 3:57–68.
Aleksandra Cichocka, Michał Bilewicz, John T Jost, Yann Mathet, Antoine Widlöcher, and Jean-Philippe
Natasza Marrouch, and Marta Witkowska. 2016. Métivier. 2015. The unified and holistic method
On the grammar of politicsor why conservatives pre- gamma (γ) for inter-annotator agreement mea-
fer nouns. Political Psychology, 37(6):799–815. sure and alignment. Computational Linguistics,
41(3):437–479.
Marco Dinarelli, Silvia Quarteroni, Sara Tonelli,
Alessandro Moschitti, and Giuseppe Riccardi. Giovanni Moretti, Rachele Sprugnoli, Stefano Menini,
2009. Annotating spoken dialogs: From speech seg- and Sara Tonelli. 2016. ALCIDE: Extracting and
ments to dialog acts and frame semantics. In Pro- visualising content from large document collections
ceedings of SRSL 2009, the 2nd Workshop on Se- to support Humanities studies. Knowledge-Based
mantic Representation of Spoken Language, pages Systems, 111:100–112.
34–41, Athens, Greece, March. Association for
Costanza Navarretta and Patrizia Paggio. 2010. Mul-
Computational Linguistics.
timodal behaviour and interlocutor identification in
Oswald Ducrot. 1995. Les modificateurs déréalisants. political debates. In International Workshop on Po-
Journal of pragmatics, 24(1-2):145–165. litical Speech, pages 99–113. Springer.
Isabella Poggi. 2005. The goals of persuasion. Prag-
Francesca D’Errico, Isabella Poggi, and Laura Vincze.
matics & Cognition, 13(2):297–335.
2010. Discrediting body. a multimodal strategy to
spoil the others image. In International Workshop Alessandro Portelli. 1985. Biografia di una città: sto-
on Political Speech, pages 181–206. Springer. ria e racconto: Terni 1830-1985. Einaudi.
Fabrizio Esposito, Pierpaolo Basile, Francesco Cu- Luisa Salvati and Massimo Pettorino. 2010. A
tugno, and Marco Venuti. 2015. The Comp- diachronic analysis of face-to-face discussions:
WHoB Corpus: Computational construction, an- Berlusconi, fifteen years later. In Interna-
notation and linguistic analysis of the white house tional Workshop on Political Speech, pages 65–74.
press briefings corpus. Proceedings of CLiC-it. Springer.
Marco Guerini, Danilo Giampiccolo, Giovanni Martijn Schoonvelde, Anna Brosius, Gijs Schumacher,
Moretti, Rachele Sprugnoli, and Carlo Strapparava. and Bert N Bakker. 2019. Liberals lecture, con-
2010. The new release of Corps: A corpus of politi- servatives communicate: Analyzing complexity and
cal speeches annotated with audience reactions. In ideology in 381,609 political speeches. PloS one,
International Workshop on Political Speech, pages 14(2):e0208450.
86–98. Springer.
Raffaele Simone. 1990. Fondamenti di linguistica.
Alan Henderson, Frieda Goldman-Eisler, and An- Laterza Bari.
drew Skarbek. 1966. Sequential temporal pat-
terns in spontaneous speech. Language and Speech, Rachele Sprugnoli, Giovanni Moretti, Sara Tonelli, and
9(4):207–216. Stefano Menini. 2016. Fifty years of european his-
tory through the lens of computational linguistics:
Hiroko Itakura. 2001. Describing conversational dom- the de gasperi project. IJCOL, pages 89–100.
inance. Journal of Pragmatics, 33(12):1859–1880. Deborah Tannen. 1989. Interpreting interruption in
Maria Koutsombogera and Harris Papageorgiou. 2010. conversation. Gender and discourse, pages 53–83.
Multimodal indicators of persuasion in political in- Khiet P. Truong. 2013. Classification of cooperative
terviews. In International Workshop on Political and competitive overlaps in speech using cues from
Speech, pages 16–29. Springer. the context, overlapper, and overlappee. In Proceed-
ings of INTERSPEECH.
Klaus Krippendorff. 2018. Content analysis: An intro-
duction to its methodology. Sage publications. Miriam Voghera. 2001. Teorie linguistiche e dati
di parlato. In Sornicola R. Stromboli C. Albano
Willem JM Levelt. 1983. Monitoring and self-repair Leoni F., Stenta Krosbakken E., editor, Dati empirici
in speech. Cognition, 14(1):41–104. e teorie linguistiche, volume XXXIII of Congresso
Internazionale di Studi della Società di linguistica
James R Lindsley. 1975. Producing simple utterances:
italiana, pages 75–95., Roma. Bulzoni.
How far ahead do we plan? Cognitive Psychology,
7(1):1–19. Valentino Zurloni and Luigi Anolli. 2010. Fallacies as
argumentative devices in political debates. In Inter-
Ferdinando Longobardi. 2010. Linguistic factors in national Workshop on Political Speech, pages 245–
political speech. In International Workshop on Po- 257. Springer.
litical Speech, pages 233–244. Springer.