Mediaforall

Round Table 3: Quantifying Audiovisual Translation Research
Convenor: Frederic Chaume
Speakers: Maria Freddi - Catalina Jimenez Hurtado - Dimitra Kalantzi - María Pavési - Cristina Valentini

ABSTRACTS

Maria FREDDI
Studying AVT through the corpus and the database: continuity and variation across translations

Recently, translation studies have benefited from the use of corpora, both parallel and comparable, aiding the study of translation specificities and tendencies (Baker 2004) or universals such as transfer, simplification and explicitation (Laviosa 1998; Mauranen and Kuijamaki 2004), and facilitating comparative and contrastive language studies (corpus-based translation studies, CTS). AVT research, in particular, has only just started to look to corpora as a resource for the study of recurrent translation solutions, alignment and disalignment with target language norms, e.g., in film dubbing, (Pavesi 2005, 2008, 2009; Romero-Fresco 2006, 2009; Bruti and Pavesi 2008; Valentini 2008; Heiss and Soffritti 2008), translators variability (Freddi 2009) and creativity.
The proposed paper discusses some results from a state-of-the-art research project which has been carried out at the University of Pavia to compile an annotated parallel corpus of both American and British films released in the period between 1995 and 2005 and their Italian dubbed versions in order to study features of dubbed language. The Pavia Corpus of Film Dialogue (Freddi and Pavesi 2009) consists of the orthographic and prosodic transcriptions of films all characterised by face-to-face conversation in daily contexts and, as a guarantee of their status and diffusion within the cultures that produced and consumed them, successful both with the critics and the general public.
The dialogues so transcribed, and with the original and the translation aligned, are related to the contextual information that has been encoded in the corpus thanks to a database structure thought out to facilitate observation of continuity patterns that cut across individual variation as well as idiosyncrasies and individual behaviour in the linguistic mediation process. Findings from the Pavia Corpus are chronologically updated through data collected in the years following 2005 to see if the same trends are observable and why.
Such empirical approach takes up the suggestion for context-sensitive analyses of translations coming from different directions from within CTS, especially Baker 2004, but also AVT, particularly Díaz-Cintas 2004. It is argued that quantitative information resulting from observation of recurrent translators’ behaviour can help pinpoint linguistic and translational areas that are problematic, improve translation quality, develop translators’ awareness and thus foster alignment with acceptability standards.

Catalína JIMÉNEZ HURTADO et al.
Multisemiotic and multimodal corpus analysis: an application to audio description

The aim of this paper is to investigate a method to apply the traditional mechanisms of corpus linguistic analysis to multimodal corpus linguistics (Baldry & Thibault (2004). Specifically, the first part of this paper will be concerned with the transformation that corpus linguistics has undergone over the past decades in the research and analysis of audio description, as well as discuss the compatibility of corpus-based methods with the research questions posed in text linguistics and its different subfields (coherence, cohesion…).
In the second part of the paper, we will propose a way to integrate the analysis of the different recurrent semiotic patterns that are present in an audio described film. This attempt includes, on the one hand, the formulation and isolation of appropriate accessible visual units of analysis in order to be able to annotate them and, on the other, to link these units to the so called language-only-texts, naturally related by the audio describer in the audio description script.
Although our collected corpus includes language-only-text as the very basis of tagging techniques and methodology, we relate this linguistic information to the grammar of the visual semiosis codified in film scenes, as well as the other semiotic structure of the film, namely its narratology.
Following Salway (2007) and Ädel & Reppen (2008) we do believe that there is a way to consider the possibility of tagging a multimodal corpus and extracting different types of information.
Researchers concerned with discourse phenomena have always been searching for solutions to include the context and co-text interpretation in their analysis. Audio description provides an excellent tool to start with it.
Projects involved: TRACCE (Code: SEJ2006-01829/PSIC) and AMATRA (Code: T07-SEJ-2660) research projects (funded by the Spanish Science and Education Ministry and the Andalusian Government, respectively) analyze the specific resource of the audio description script in a corpus of 200 scripts from AD films.

Bibliography:
Ädel, Annelie and Randi Reppen (2008): Corpora and Discourse. The challenges of different settings. Amsterdam: John Benjamins.
Alwood, Jens. (2008) “Multimodal corpora”. In Lüdeling, Anke & Merja Kytö (eds.), Corpus Linguistics. An international Handbook. Berlin, New York: Walter de Gruyter, 207-225
McNamara, Danielle.S., & the CSEP lab (2003). “Coh-Metrix: Automated cohesion and coherence scores to predict text readability and facilitate comprehension.” Annual project report submitted to the Institute of Education Sciencesto predict text readability and facilitate comprehension.
In http://csep.psyc.memphis.edu/mcnamara/vita.htm
Nivre, Joakim. (2008) “Treebanks”. In Lüdeling, Anke & Merja Kytö (eds.), Corpus Linguistics. An international Handbook. Berlin, New York: Walter de Gruyter, 225-241
Salway, Andrew. (2007), ‘A Corpus-based Analysis of Audio Description’, in Díaz Cintas, Orero and Remail (eds.), Media for All: Subtitling for the Deaf, Audio Description and Sign Language, Amsterdam – New York, Rodopi, pp. 151-174 Salway, A. (2007), ‘A Corpus-based Analysis of Audio Description’, in Díaz Cintas, Orero and Remail (eds.), Media for All: Subtitling for the Deaf, Audio Description and Sign Language, Amsterdam, New York: Rodopi, 151-174
Taboada, Maite., Kimberly Voll and Julian Brooke (2008) “Extracting Sentiment as a Function of Discourse Structure and Topicality”. School of Computing Science Technical Report 2008-20.

Dimitra KALANTZI
Subtitling for D/deaf and HoH people and corpus-based research

There is no doubt that Subtitling for the D/deaf and HoH (SDH) offers an invaluable service to millions of D/deaf and HoH viewers providing them with access to audiovisual media broadcast on television and increasingly in cinemas, DVDs and on the Internet. However, very little research has been dedicated to describing current subtitle output; for instance, although verbatim vs. edited subtitling is a much discussed issue in SDH, we still do not know to what extent current subtitles are verbatim and to what extent they are edited. In this light, large-scale corpus-based research is deemed essential, as not only will it allow us to better map the world of SDH, but it will also constitute the first step towards determining the needs and preferences of D/deaf and HoH viewers and ultimately in providing consistently high-quality subtitling.

In addressing the issue of this paper, I will:

Point out the need to adapt current corpus typology in order to accommodate typical intralingual SDH corpora, i.e. corpora consisting of texts in one language and their subtitles in the same language
Suggest features to measure and metrics to use in order to describe current subtitles
Underscore the need for standardisation: this includes among other things determining how to measure subtitle rate and the need to identify a suitable metric, determining how to measure deletion and defining/agreeing on instances of poor segmentation within subtitles
Report some initial findings on the basis of my research and suggest relevant hypotheses.

María PAVESI
What corpora can tell us about the language of dubbing

This contribution will focus on the language of AVT, conceived as a type of translated language worth investigating in itself, and in the search for regularities at various linguistic, generic and translational levels. Although I will be speaking about dubbing, I believe that my observations could be easily applied to the language of other types of AVT such as subtitling, audio description or televised interviews.
I will start from the widely shared belief that metalinguistic introspection, insight and individual appraisal are not sufficient to gain full access to the regularities of linguistic and translational behaviour. To reach valid and reliable generalizations on dubbed language, therefore, we need to support statements based on experience or single case-studies with quantifications on large samples of dubbed texts, in line with research on other translated languages (cf. Corpus-based Translation Studies). It is only through analyses of wide collections of primary data that we can identify systematic patterns of translational outcomes, infer norms of translational behaviour and assess the role of the universals of translation in this specific transfer activity.
Although thorough profiles of the language/languages of dubbing are needed to compare with recent findings on original audiovisual languages, the investigation of individual features or clusters of related features can tell us quite a lot about the nature of dubbed language, its place in the target community’s sociolinguistic repertoire, its relationship with the source language texts and its functions in audiovidual translation. I will briefly review some of the investigations carried out on the language of dubbing, and distinguish between onomasiological and a semasiological approaches depending on whether the starting point for analysis are the communicative functions performed in the fictional dialogue (e.g. speech acts) or the language structures found in film language (e.g. discourse markers, conversational routines). Following the latter procedure and relying on the Pavia Corpus of Film dialogue, I will examine the results on two clusters of features – interjections and personal pronouns – in dubbed language, with specific reference to Italian translated from English. A careful and flexible use of translated, original and reference corpora will be shown to help assess the degree of naturalness of dubbed texts, define the role of interference in AVT and highlight the interactional dimension of film dialogue.

Cristina VALENTINI
A quantitative study of communicative acts and situations in dubbing: how much do they tell about film translation and discourse?

The present paper sets out to illustrate a possible application of corpus-based methods to the field of AVT drawing on the experience of Forlixt 1, the Forlì Corpus of Screen Translation. Particularly, the present study will aim at showing how the exploitation of a multimedia, annotated, reciprocal corpus can help cast a new light not only on the analysis of translation strategies, but also on the way original film dialogues are built as far as narrative aims, plot structuring, message conveyance, and, ultimately, genres’ conventions, are concerned.
First we will exemplify the extent to which the annotation of some pragmatic phenomena (i.e. communicative acts and situations) can help contrastively point out patterns of recurrent translational behaviour in a corpus of Italian dubbed products from French. Next we will attempt at explaining such results against other significant linguistic features or specific semiotic constraints identified. In so doing, we will show how it is possible to elicit well-founded quantitative results together with significant context-dependant qualitative solutions by applying a corpus-based approach to AVT.
Second we will illustrate how the frequency and distribution of such categories in the original corpus can also usefully help emphasize some intrinsic features of cinematic and television’s dialogues, in terms of communicative verisimilitude, observance of genre-related conventions, and explicitness of film discourse. Such results may eventually provide the case for reviewing the traditional approach to the investigation of dubbed language, which may be less influenced from the translation process than from inherent features of original screenplay writing.