Forlixt 1 : A multimedia database for AVT research

By Cristina Valentini & Sabrina Linardi (University of Bologna)

Abstract & Keywords

The present article begins with an overview of technical and methodological considerations on the construction of the Forlì Corpus of Screen Translation (Forlixt 1) notably including data collection and digitisation, data segmentation and alignment, as well as data annotation. The second part of this contribution will set out to describe the database’s query functions, providing examples mainly selected through tags assigned to specific occurrences of diatopic variation. In so doing, both advantages and disadvantages deriving from a scientific exploitation of film corpus data in contrastive translation studies will be outlined for discussion.

Keywords: multimedia translation, multimedia corpora, corpus annotation, audiovisual translation, dubbing

©inTRAlinea & Cristina Valentini & Sabrina Linardi (2009).
"Forlixt 1 : A multimedia database for AVT research"
inTRAlinea Special Issue: The Translation of Dialects in Multimedia
Edited by: M. Giorgio Marrano, G. Nadiani & C. Rundle
This article can be freely reproduced under Creative Commons License.
Permanent URL:

1. Introduction[1]

The construction and use of multimedia corpora has been advocated for a while in the literature as one of the expected future application fields of Corpus Linguistics (Bernardini 1999, 2000; McEnery et al. 1996, 1997; Baldry et al. 2001; Baldry et al. 2004). In 2003, in response to similar challenges, a research group on audiovisual translation (hereafter AVT) of the Department of Interdisciplinary Studies in Translation, Languages and Cultures (SITLeC) of the University of Bologna, set out to design a multimedia database for the collection and study of data pertaining to film translation, subsequently baptized with the acronym Forlixt 1, the Forlì Corpus of Screen Translation[2].

This contribution will first attempt a brief overview of the current state-of-the-art in the conception and development of multimedia corpora, and a description of the theoretical and scientific principles on which it is possible to ground the creation of such corpora. Second, a detailed account of data collection, digitisation, segmentation, and annotation methods and steps will be provided, with a view to underline some of its most innovative technical aspects and contents. Finally, the third part will illustrate how the database works and its multiple possible applications, building on examples of dialects and regiolects embedded in film discourse, with particular regard to AVT research.

2. Multimedia Corpora

Since 1971 when Katharina Reiss provided a first theoretical definition of the concept lying beyond the term audio-media, emphasising the importance of the interdependence of the various semiotic modalities for translation (Snell-Hornby 1996), the term multimedia has been applied to a variety of situations where the exploitation of several media is involved to convey meaning. In particular, the expression multimedial, or rather audio-visual translation (Gambier 2003, 2004) implies referring to a translation process where the visual, verbal and auditory levels of a text have to be considered as a whole, since they all take part into the construction of meaning. Moreover, by the end of the last century, new forms of communication (CD ROM, DVD, hypertexts, websites, e-commerce, e-books, etc.), which often include multilingual and multimedia texts, have been gaining ground. This has led scholars in the field of Translation Studies to come to grips with these new text types. Nevertheless, it is only from the 90s that the academy started to seriously commit itself to multimedia (film and TV) translation.

The original idea behind the conception of a multimedia corpus can be traced back as early as 1994 when a research group working at the Advanced School of Modern Languages for Interpreters and Translators (SSLiMIT) of the University of Bologna, through the organisation of a series of conferences and the publication of several contributions on various aspects of screen translation (Baccolini et al. 1994; Heiss et al. 1996; Bosinelli et al. 2000; Chiaro et al. 2008), were among the first to meet the challenges issued by certain branches of Corpus Linguistics (McEnery et al. 1997; Jones et al. 1998) by advocating a possible application of CL instruments and methods to this field. A new alternative was put forward “to draw up alongside the written text, original soundtrack and images, linking them to the textual corpus by means of a series of hypertextual links” (Bernardini 2000: 306)[3]. Accordingly, the researcher would be allowed to exploit the corpus in its more accessible format (the written mode) with the possibility to refer to the original form of texts (audio-visual) whenever interpretation difficulties or peculiar research needs should arise.

These first theoretical considerations on multimedia corpora’s construction, however, still viewed the possibility to access the full multimedia content only as an option to further contextualize empirical findings obtained from the exploration of traditional written data (i.e. transcripts of film dialogues). On the other hand, from a methodological point of view, they implied that it was not correct to process multimedia data on the basis of their transcripts, in that this will necessarily lead to a flat, incomplete description not taking into account specific features, like prosody, intonation, the link existing between the verbal and the non-verbal aspect of communication (e.g. gestures and facial expressions), as well as situational constraints and characteristics of film editing, which notably represent a constraint for dubbing practitioners. Although the need to go beyond the application of mere linguistic approaches to multimedia translation, adopting instead a comprehensive semiotic perspective, inform the majority of studies recently published (see, in particular, Chaume 2004), academic research has to date failed in elaborating adequate theoretical models capable of taking into account all communicative models involved (Nergaard 2000; Cattrysse 2001). It seems, moreover, that the majority of these innovative studies are mainly case-studies or investigations that mainly focus on very specific aspects, without actually leading to the compilation of structured multimedia collections of significant amounts of data allowing to draw empirically-sound results and replicable generalizations (Gambier and Gottlieb 2001).

In response to similar observations and suggestions, a hypothesis has been formulated to put together a database destined to AVT research and teaching around a corpus of film material. The main aim of the Forlixt 1 project is notably to allow the adoption of a global corpus-based approach to AVT, eventually leading to the definition of more accurate models of analysis. The only other experience in such direction, to our knowledge, comes from systemic-functional linguistics, and in particular, the multimodal analysis theory (Thibault 2000). Since 2000, multimodal transcription has actually provided the scientific basis for the conception of a Multimodal Concordancing Authoring system (MCA) developed by a joint research team from the Universities of Pavia and Trieste (Baldry 2001 et al.; Taylor 2003, 2004; Baldry et al. 2004; Baldry 2006)[4]. Its main aim is to study the synchronisation of meaning-making resources deployed in film texts “with a view to understand the differences and similarities between genres and establish recurrent patterns of meaning making” (Baldry et al. 2004). Similarly, one of the specific aims of Forlixt 1 is to find recurrent translational patterns in film texts by examining the interdependence between language and other semiotic modalities. However, the investigation of such other semiotic aspects is instrumental in pushing further the study of the verbal correlate. In this sense, images and sounds are regarded as useful and sometimes unavoidable elements that help contextualise and clarify speech, but the wording remains nonetheless the main focus of research in Forlixt 1. By contrast, Forlixt 1 is mainly a translation-oriented corpus with a multilingual architecture, as hereinafter pointed out.

3. Forlixt 1: Corpus Construction

The Forlixt 1 corpus is hypothetically composed of multiple sub-corpora, parallel or comparable, which may be created, queried and customised according to different needs by means of a special filter page working as a gateway for accessing textual and multimedia resources. By parallel corpus we traditionally mean a collection of texts in language A and their translations into language B, C, D, etc. (Teubert 1996). Although Forlixt 1 presently contains only Italian, French and German-speaking countries’ products, the software tool has been designed in such a way as to welcome the collection of an indefinite number of languages. Moreover, following Teubert’s definition, Forlixt 1 comprises scenes of films and transcripts aligned on a scene-by-scene basis so that occurrences can be retrieved along with their translation, from language A into language B, but also from language B into language A. Thus Forlixt 1 is also a multidirectional database. Finally, it can also be defined as comparable in that it is made up of original films in both languages, A and B, which can be ascribed to the same film genre and are therefore similar in terms of their contents and narrative structure[5]

Since corpora are by definition compiled in order to be representative of a particular language or variety, and their size, contents and organization may vary according to the purpose they are required to fulfil (Ulrych 2001), it is necessary to accurately define the purposes and uses of the present corpus before discussing it in terms of its representativeness. Forlixt 1 has been designed first and foremost to be an archive for storing films and their tran(scripts) and to make them accessible both in their original form and through an index of pre-established categories, in order to satisfy academic research needs on AVT. As such, the corpus is representative of the linguistic variety characterising traditional “written to be spoken as if not written” film discourse (Gregory et al. 1978) and translated discourse, which can either take the form of a peculiar spoken (dubbing, audio-description, commentary, …) or, again, written (subtitling) variety. Film dialogues trying usually to reproduce/mime spontaneous speech eventually create a language which prototypically finds itself in between writing and speech. This feature contributes to raising specific questions about methods of transcription of speech, use of punctuation, superimposition of the notion of utterance/sentence/conversational turn and specific problems of speech segmentation, similar to those addressed by the C-ORAL-ROM project (Cresti and Moneglia 2005). Another specific issue deals with the transcription of oral diatopic varieties that in many cases lack written standardization. Although Forlixt 1 is not to be intended as a database specifically conceived for the study of dialectal and regiolectal varieties, specific categories help find occurrences of the latter in the corpus provided the user is made aware of limitations related to such a specific application, particularly concerning access to non-standardized (non-phonological) transcriptions of dialects.

Forlixt 1 is currently made up of 75 films and TV series’ episodes – 12 of which are original Italian, 9 original German and 16 original French - including the complete transcription of film dialogues. The corpus currently amounts to about 115 hours of fully transcribed, synchronised and annotated audiovisual material[6].

3.1 Data collection and digitisation

The criterion applied in the selection of material was first of all availability, intended as the likelihood to find films in the home video markets of countries in which such films were distributed in the original shooting language, as well as of those countries in which a dubbed version of them was released for home vision[7]. The first step in the building up of the electronic corpus has accordingly required the digitisation of DVD or VHS material. A specific video capture software - Pinnacle Studio PlusTM version 10.5 - has been employed to record movies into standard Mpeg2 video files. Such video files have been subsequently converted into a lighter format, i.e. Windows Media Video 8 for Local Area Network (768 kbps), the choice of this format being dependent on the type of server selected for storing video data, which was Microsoft Media Services TM.

3.2 Data segmentation and alignment

The retrieval of a part or the whole of a moving image alongside with its textual information from a systematically organised collection depends upon how each moving image and text, and parts thereof, are segmented and described: “intuitive and innovative forms of video retrieval, browsing and reuse require video data to be made into a structured medium by analysis and description of its content” (Salway 2005: 1). In order that films can be accessed, it is necessary to segment the audio-visual stream into video intervals, e.g. scenes, and associate them with the corresponding text fragments. To do so, we adopted Windows Movie Maker TM version 5.1, a video editor that, provided a film is imported in a collection creating a specific project file, permits the visualization of the complete filmic sequence and the identification of the time code marking the beginning of each line. This information represents the smallest information the software can process. Next, lines, along with other relevant information (name of the character, language of the line) require to be imputed in a dedicated database specifically created for storing multimedia data.

To build up a suitable architecture for content organization and management, a specific software tool whose function was similar to that performed by MovieDB, currently at its 11th release, has been developed. This software is essentially a relational database based on a SQL architecture permitting the structuring and management of data in the form of collections of tables that are logically associated to each other by means of shared attributes. The first two releases of the database were stand-alone access versions and served to pilot the software. An SQL server version of the database has been subsequently designed using Microsoft SQL server 2000, whereas the first access-based interface has been retained for data entry activities thanks to the implementation of a pass-through technology, allowing the entering of data from a remote workstation (figure 1).

Image 1: General Overview


A manual data entry approach has been adopted. The main entry window of MovieDB (figure 2) allows for the entering of general information for each film: title of the film, language, original or dubbed version, year of production/dubbing, names of film actors, directors, film genre. This information will be used to browse the corpus and will function as a preliminary filter for restricting the domain under investigation. Moreover, this page allows the user to enter the link to the windows media video file of the film, formerly stored into the above-mentioned Media Services server, as well as the link to the original version of the film, whenever the film in question is a dubbed version.

Image 2: General Information


The three buttons (dialogues, subtitles and scenes) in the lower part of the page give access to other pages and functions specifically designed to perform tasks concerning video and text segmentation, and multimedia annotation. Figure 3 shows the first sub-task for the current film session, which mainly consists in segmenting both the audiovisual and text file into smaller units (lines). This is possible by selecting the initial time code of the line, that is the exact point in time where the character starts to utter his/her line, and entering it together with its corresponding transcription. Hence, each single transcribed line is associated in the database with the corresponding audio-visual shot, allowing the software to process and retrieve simultaneously both pieces of information. The added value of such an approach has been recently confirmed by MCA developers who have emphasized the extent to which it is important “to access texts in an in vivo form that provides access to the audio and video tracks and maintains their relationship intact” (Baldry 2004: 24-25).

Image 3: Dialogues


Concerning this last point, we have generally agreed that transcriptions must exactly mirror the sequence of words uttered by the speaker, to which it must be added, as far as possible, all non-lexical elements existing in the acoustic signal (truncations, hesitations). This also applies to the transcription of some Italian dialects and regiolects for which there are no clear or widely recognized standards. They have been entered into the database trying to provide an orthographic transcription as close as possible to a verbatim speech. In other cases, instead, they have been normalized. This has been the case of the film Nordrand (1999, Barbara Albert)[8] .

The second sub-task involves the chunking of the video file into scenes by assigning an initial time code and a final time code (figure 4). This can be done either by recalling the time codes from a list of time codes imputed during the first level of data entry, or entering a new initial and/or final time code. Such intervals are processed by the software for the alignment of scenes and parts of film dialogues. To do so, however, each scene has to be matched with the corresponding scene in the original language.

Image 4: Scenes


3.3 Data annotation

The last step consists in annotating each relevant scene by assigning it one or more attributes, or labels, which need to be recalled from a specially devised list of lingua-pragmatic categories (figure 5). The current system is built upon a multi-level tree of hybrid categories (labels) that can be browsed to retrieve hit lists pertaining to specific lingua-pragmatic and socio-cultural phenomena (Heiss 2005; Heiss and Soffritti 2008; Valentini 2008). For instance the first level of indexation comprises communicative situations, communicative acts, settings, cultural specificities, and linguistic varieties. Each of these macro-level categories comprises at least one more level of analysis with labels ever more specific to designate specific lingua-pragmatic phenomena.

Image 5: Multimedia Annotation Tree


Methodologically, to accurately annotate film scenes it is first necessary to carry out a qualitative analysis on the filmic text. With this respect, categories used can be ascribed to two different groups: on one hand, we have a list of formal and discrete categories, that is to say commonly accepted and universally shared. For instance names of places, diseases, famous people, et cetera; on the other hand, we also have to cope with a list of less formalized categories. Even though a working definition is established to uniformly carry out the annotation, they nonetheless present some blurred areas of juxtaposition with nearby categories. This applies, for instance, to communicative acts and situations.

Another specific problem concerns the level of detail of the analysis and particularly the level of accuracy in describing certain phenomena. With this respect, two principles have been retained. On one hand, the principle of the semiotic relevance of a given phenomenon, that is to say the importance and density of the information conveyed. This appears extremely relevant if we are to adopt a mono-linguistic and mono-cultural approach, concerning as well the question of the annotation of non-verbal elements (special effects, signs, graphical elements, cultural objects, music, etc.) that actively contribute to the making of a film’s meaning. On the other hand, from a contrastive point of view, a need emerges to identify critical points and elements witnessing a cultural and linguistic clash in the original and dubbed version, as well as phenomena that are likely to bear witness to specific translation strategies (e.g. omission, generalization, replacement, etc.). Be that as it may, the system allows for the retrieval of non-annotated data by associating the various search systems and combining different levels of micro and macro analysis (Heiss and Soffritti 2008).

4. On Possible Applications of Forlixt 1 in AVT research[9]

Forlixt 1 represents an innovative tool offering the possibility of accessing multilingual audiovisual material in a contrastive way and inferring generalizations about dubbing strategies and patterns of AV constructs. It offers an ideal basis on which to ground empirical and quantitative research, thus helping establish a scientific approach in the discipline of audiovisual translation studies, as it has been discussed in other papers focusing on Forlixt 1 (Heiss and Soffritti 2005, 2008; Valentini 2007, 2008). Conversely, Forlixt 1 can be used as a useful support in AVT teaching. As a matter of fact, thanks to its features, the database is a valid tool to be used as a component in the training of audiovisual translators, in that it can help develop traditional linguistic, but also communicative and cultural skills (Valentini 2006). Besides, the study of problems pertaining to film translation represents an innovation in the field of translation research presenting specific challenges if it is to be exploited in teaching (Heiss 2000). Presently, data can be analyzed through four specific search systems, namely the full-text search, the guided search or search by attributes, the advanced search and the combined search. A specific aim of this section will be to show how the database can be fruitfully exploited to substantiate research on the translation of diatopic varieties occurring in fictional reproduction of day-to-day language.

4.1 Full-text search

The database can be searched via an html interface either by applying a possible preliminary filter or directly starting a full-text search or a guided search [10]. The free-text modality allows the user to look for words or string of words, i.e. a linear sequence of words. Queries can also be restricted by sub-domain, namely selecting one of the given options: dialogues or subtitles or dialogues and subtitles. It is also possible to enter a query expression in between inverted commas, using the logic operator asterisk (*) to replace one or more characters. The software provides the results in the form of a list of occurrences of the word or string of words queried accompanied by its surrounding textual context, namely the complete line in which the query string appears. For each occurrence it also provides some general information concerning the film (original or dubbed version, the film title, the name of the character uttering the line, and the language of the film). Finally, this page contains the hypertextual link to the scene associated with each line. A new page allows the user to play the video scene including the line queried and to compare it with its orthographic transcription. By selecting the target language from the language menu bar, a bilingual comparative analysis is possible. The software retrieves the parallel text in the language selected together with the associated scene.

An example of a possible application of the full-text search will demonstrate how useful this tool can be in audiovisual translation research. Knowing that the database contains the Italian and German version of two films of the Emilian film director, Pupi Avati, we will try to find examples of the dialect spoken in the Northern part of the Emilia-Romagna region, namely Emilia, entering the term socci*, a quite vulgar but frequent dialectal expression of surprise in order to see which translation strategies have been adopted. The film Festa di laurea (1985, Pupi Avati) presents four occurrences of this term. The film is set at the end of World War II. A middle-aged working class man, who has just started his own bakery, has agreed to prepare a run-down beach house with an overgrown garden for a graduation party on an extremely tight schedule, together with his son and two young employees.


In this scene, one of the employees brings a radio in the house provoking a reaction of positive surprise in the protagonist’s son expressed by the dialectal term socci. In the German version, even if the expression klasse (peachy) is flat and standard, the emphasis the actor puts on the word radio and the boy’s naïf attitude confer a meaning of positive surprise.


In this scene a man enters a room, sees all the expensive presents for the graduated girl and cries out: Soccia che bella roba! (Wow, what beautiful things!). In the German version, the exclamation Donnerwetter is used both to express imprecation and informal admiration. It represents a convincing translation although the sentence in the dubbed version is difficult to be heard.


Here the two employees, knowing that the radio is stolen, are worried about the arrival of a man they suppose to be a policeman. In this case the dialectal term soccia (Oh my God) has been omitted in the dubbed version, so that the speaker seems less worried than in the original.


Here, at sunrise after the party, the two employees are walking along the street when the protagonist and his son, who were looking for them, arrive by car. The protagonist’s son’s exclamation Soccia, è due ore che vi cerchiamo! (Damn, we have been searching for you for two hours) expresses his tiredness, his apprehensiveness and in some way his desire to be considered a boss by them. The omission of soccia and the lack of emphasis in this sentence results in a flat dubbed version.

On top of that, the film presents an incorrect translation in a scene which follows that of example 3. In the original version, the protagonist’s son asks Oh, ma quella radio sei sicuro che l’hai riportata indietro? (Are you sure that you really brought back that radio?) revealing his fearful attitude and mistrust. On their part, none of the employees who stole the radio seem so sure that the radio has really gone back to its owner, and one of them answers C’ha pensato Dario… (Dario did it…) with a kind of uncertainty in his voice underlined by his body language and a gesture of his hands. On the other hand, the back translation of the dubbed version is You must in any case bring back that radio (Also, dieses Radio müsst ihr auf jeden Fall zurückbringen). The act of request of the original scene has been changed into a sort of exhortation; the past tense is changed into a present, the second person singular into a second person plural. The employee answers with a tone of superiority and casualness Dario intended to do it anyway underlined by his body language which in this case could be interpreted as if he had intended to say Don’t get flustered. In so doing, these scenes illustrate the tendency in dubbing to standardize varieties employed or not to translate them at all (Nadiani 2004).

4.2 Guided search

The guided query allows for the exploration of the corpus from a pragmatic and linguistic point of view, according to a hybrid set of criteria corresponding to the list of attributes used for tagging the corpus. In order to start, for instance, a research on the use of dialects, one can first restrict the corpus by applying the filter original and Italian and then select the categories linguistic varieties>regional and social varieties>dialect from the tree. The category dialect was used to annotate parts of dialogues which vary in form, pronunciation and grammar from the standard variety and are typically spoken in single cities or small areas. If it is not possible to precisely locate the city or region in which a certain dialectal variety is spoken, because specific geographical occurrences may be ascribed to a wider area (i.e. merely concerning pronunciation and prosody), the category applied is regiolect.

4.3 Advanced search

The advanced search consists in a guided search carried out on the results of a previous guided search of up to eight possible levels of refinement, thus permitting the user to combine different aspects of the analysis. Dialect is used in films for several reasons and there are different ways to deal with it in translation. Given the fact that dialectal elements are seen as generally evoking a funny effect (Schröder and Stellmacher 1989; Schröder 1995), one could search by resorting to this technique whether there is a connection in the film La vita è bella (1997, Roberto Benigni) in the use of dialect and verbally expressed humour, a label that is employed in the corpus to indicate the presence of paronomasia, puns, wordplays. By restricting the corpus to this film through the filter page, it is possible to start a guided search, choosing the category dialect>Tuscan and then applying to the results of this query the option search among the results and again the category verbally expressed humour. This search eventually reveals that in 50% of the scenes the two attributes are connected (see figure 6).

Image 6: Advanced search


In the German film, Zuckerbaby (1985, Percy Adlon) the regiolect and dialect of the Bayern region of Germany are used in many circumstances and by different characters. The use of dialect in this film plays different functions, for instance it underlines the colloquial and informal component of spoken language and acts as a social marker; it defines certain characters as often happens in television programs (Herbst 1994). Dialectal expressions are very frequent in speech acts like greeting and leave-taking. Through the advanced search tool, looking for the combination of the category bayerisch and the speech act leave-taking, one can investigate how dialectal forms of leave-taking have been translated. A typical expression of leave-taking is servus. In the film Zuckerbaby this expression has been translated in most cases with the Italian greeting term ciao, which lacks any regional connotation, or with the expression ti saluto, which is not as common and spontaneous as the original. In particular, in a scene of leave-taking between two metropolitan drivers, a translation mistake occurs: the appellative Grantler (in German describing someone who is always in a bad mood) was not translated, because it was considered to be the name of the driver (Aldeghieri 2000).

By applying a combination of the two categories leave-taking and regiolect>south another translation mistake emerges: here the hobby Bürokegeln (meaning “to play bowling with colleagues after work”) has been translated with a sarcastic expression Gioco a birilli in ufficio (I play bowling in the office). As underlined by the analysis of these two scenes, some comprehension difficulties occur in the translation of this films. Although the replacement of a dialect in the source text with another dialect of the target language is not advisable (Herbst 2004), the application of other compensation strategies for dialectal and regiolectal varieties are nonetheless recommended in specific circumstances (Heiss 2000).

4.4. Combined Search

This last innovative query system allows for a combined search by running a guided search on the results of a full-text search. For instance, entering the dialectal Italian word sacciu (I know), which is used in the southern Italian regions of Sicily and Calabria, and applying the attribute facial expressions to the results of the full-text search, one can see if there is a relevant connection between the way the word is spoken and the facial expressions of the characters.

Image 7: Combined search


A typical example found is a scene taken from the comedy Mimì metallurgico ferito nell’onore (1971, Lina Wertmüller). The story, set in Sicily and Turin, is about a Sicilian dockworker who loses his job when he votes against the Mafia candidate.


In this scene, the protagonist, after a gun shot, refuses to tell the police what he has seen. Here the expression underlines what Mimì is saying. The line in the dubbed version is just one sentence with two subordinate clauses (I’m really sorry but I can’t help you because I don’t know anything) and seems less spontaneous and less convincing than the original. According to Paolinelli and Di Fortunato (2007) when dealing with dialectal or regiolectal varieties, audiovisual translation needs to focus in any case on the communicative purpose, so that often a compromise between a specific linguistic production mode and the communicative intent needs to be found.

Coming back to the expression servus which is used in South Germany as well as in Austria, it is possible to search for scenes in which it is employed to see how it has been translated as an expression of greeting or leave-taking. In the Austrian film, Nordrand (1999, Barbara Albert) the dubbed version presents two different translations of servus in the same scene. A girl greets her friend and he replies ciao, while another boy, whom the girl does not know, greets her with the expression salve, which in Italian is not as common as ciao among young people. As far as the presence of the expression salve is concerned, restricting the corpus by applying the filter original films/dubbed films the search reveals that among the 61 films contained in the database, there are many more instances of this term in dubbed films than in original versions, with 22 occurrences in dubbed films and only 10 occurrences in original ones.

5. Conclusion

This paper has dealt with the description of Forlixt 1, the Forlì Corpus of Screen Translation, an experimental database specifically conceived for audiovisual translation research. First of all, a detailed overview of corpus construction and annotation methods has been provided, emphasizing the major problems encountered in multimedia and text segmentation, alignment and annotation. The second part of this contribution has tried to discuss the four main query tools and show how each of them can be used to analyze and select data, particularly by combining textual occurrences with annotated multimedia phenomena. This has eventually enabled us to extract examples of regiolectal and dialectal varieties of Italian and German embedded in film discourse, thus contributing to laying down a new approach in the experimental study of the translation of geographical variation in fictional products across different languages. 


Aldegheri, Chiara (2000). “Il doppiaggio del film ‘Zuckerbaby’ di Percy Adlon. Analisi degli aspetti linguistici e culturali”. Unpublished dissertation. SSLMIT, University of Bologna at Forlì

Baccolini, Raffaella, Bosinelli Rosa Maria and Laura Gavioli (eds) (1994). Il doppiaggio: Trasposizioni linguistiche e culturali. Bologna: CLUEB

Baldry, Anthony and Thibault, Paul, J. (2006). Multimodal Transcription and Text Analysis. A multimedia toolkit and coursebook. London: Equinox

Baldry, Anthony (2004). “Concordancing film corpora: the role of MCA in multimodal discourse analysis” in Carol Taylor, Maria Grazia Busà and Sara Gesuato (eds.) Proceedings of the Seminar, Linguatel: Progetto telematico di lingua e mediazione interlinguistica, 9-10 May 2002, University of Padua, Padova: Unipress, 19-32

Baldry, Anthony and Beltrami Michele (2004). “Report: The MCA system: its role in the development of multimodal corpus linguistics” in: DIDACTAS: Didattica dell’analisi testuale, dei corpora, della traduzione e della sottotitolazione [online] (Last Access: June, 2008)

Baldry, Anthony and Paul J. Thibault (2001). “Towards Multimodal corpora” in: Guy Aston and Lou Bernard (eds) Corpora in the description and teaching of English. Papers from the 5th ESSE conference, Bologna: CLUEB, 87-102

Bosinelli Rosa Maria, Heiss Christine, Marcello Soffritti and Bernardini Silvia (eds) (2000). La traduzione multimediale. Quale traduzione per quale testo? Bologna: CLEUB

Bernardini, Silvia (1999). Competence, capacity, corpora. Bologna: CLUEB

Bernardini, Silvia (2000). “Corpora, traduzione, multimedialità” in: Rosa Maria Bollettieri Bosinelli, Christine Heiss, Marcello Soffritti and Silvia Bernardini (eds) La traduzione multimediale. Quale traduzione per quale testo? Bologna: CLEUB, 301- 321

Cattrysse, Paul (2001). “Multimedia & Translation: Methodological Considerations”, in: Gambier, Yves and Gottlieb Henrik (eds) (Multi) media translation: concepts, practices and research. Amsterdam/Philadelphia: John Benjamins, 1-13

Chaume Varela, Federico (2004). Cine y traducción. Madrid: Catédra

Chiaro, Delia, Heiss, Christine and Chiara Bucaria (eds) (2008). Between Text and Image: Updating Research in Screen Translation. Amsterdam/Philadelphia: John Benjamins

Cresti, Emanuela and Moneglia, Massimo (eds) (2005). C-ORAL-ROM : integrated reference corpora for spoken romance languages. Amsterdam/Philadelphia: John Benjamins

Heiss, Christine and Rosa Maria Bosinelli (eds) (1996). Traduzione multimediale per il cinema, la televisione e la scena. Bologna: CLUEB

Gambier, Yves (2003). “Screen Transadaptation: Perception and Reception” in: The Translator, 9 (2), 171-189

Gambier, Yves (2004). “La traduction audiovisuelle: un genre en expansion”  in : Meta, 49 (1), 1-11

Gambier, Yves and Gottlieb Henrik (eds) (2001). (Multi) media translation: concepts, practices and research. Amsterdam/Philadelphia: John Benjamins

Gregory Michael and Susanne Carroll (1978). Language and Situation: language varieties and their social context. London: Routledge & Kegan Paul

Heiss, Christine (2000). “La traduzione filmica come pratica didattica” in: Bernardini Silvia, Bosinelli Rosa Maria, Heiss Christine and Marcello Soffritti (eds) La traduzione multimediale. Quale traduzione per quale testo? Bologna: Clueb, 183-196

Heiss, Christine (2004). “Translatorische Aspekte der Filmsynchronisation Deutsch/Italienisch/Deutsch”, in: Moderne Sprachen, 48(2), 39-59

Heiss, Christine and Soffritti Marcello (2005). “Parallelkorpora ‘gesprochener Sprache’ aus Filmdialogen? Ein multimedialer Ansatz für das Sprachenpaar Deutsch-Italienisch”, in: Korpus
Linguistik Deutsch - synchron, diachron, kontrastiv
. Tübingen: Niemeyer, 207-217

Heiss, Christine and Soffritti Marcello (2008). “Forlixt 1: The Forlì Corpus of ScreenTranslation: Exploring Micro-structures”, in: Chiaro Delia, Heiss Christine and Chiara Bucaria (eds.) Between Text and Image: Updating Research in Screen Translation. Amsterdam/Philadelphia: John Benjamins, 51-62

Herbst, Thomas (1994). Linguistische Aspekte der Synchronisation von Fernsehserien. Phonetik, Textlinguistik, Übersetzungstheorie. Tübingen: Niemeyer

Jones, Randall L. and Michael D. Bush (1998). “Accessing sounds from a spoken corpus of German: A feasibility study” in: Teaching and language corpora 1998 (pre-conference volume). Oxford: Oxford University Humanities Computing Unit, 96-98

McEnery, Tony and Andrew Wilson (1996). Corpus Linguistics. Edinburgh: Edinburgh University Press

McEnery, Tony and Andrew Wilson (1997). “Multimedia corpora” in: Barbara Lewandowska-Tomaszczyk and Patrick James Melia (eds) Practical applications in language corpora – Proceedings. Lodz: University Press, 24-33

Nadiani, Giovanni (2004). “Dialekt und filmische Nicht-Übersetzung – Der einzig möglich Weg?”. Dialekt-Übersetzung und Dialekte in Multimedia, hrsg. Von Irmeli Helin, Frankfurt am Main: Peter Lang, 53-74.

Nergaard, Siri (2000). “Un approccio semiotico alla traduzione multimediale” in: Bollettieri Bosinelli, Rosa Maria, Heiss, Christine, Soffritti, Marcello and Bernardini, Silvia (eds) La traduzione multimediale. Quale traduzione per quale testo? Bologna: Clueb, 431-451

Paolinelli, Mario and Eleonora Di Fortunato (2007). “L’adattamento dei dialetti: un non-problema” Paper presented at the international conference “Mulimedialectranslation 2007”, Forlì, 10-12 May 2007, SSLMIT: University of Bologna at Forlì

Salway, Andrew and Ian Xu (2005). “Navigating Stories in Films”’, Dept. of Computing Technical Report CS-05-04, University of Surrey [online] (last access: 10/02/2008).

Schröder, Martin (1995). Humor und Dialekt: Untersuchung zur Genese sprachlicher Konnotationen am Beispiel der niederdeutschen Folklore und Literatur. Neumünster: Wachholtz. (Göttinger Arbeiten zur niederdeutschen Philologie, 14)

Schröder, Martin and Stellmacher, Dieter (1989). “Das Verhältnis von Humor und Dialekt: ein Problem der Sprachbewertung”, in: Zeitschrift für Dialektologie und Linguistik, 56 (S), 171-181

Snell-Hornby, Mary (1996). “‘All the world’s a stage’: Multidimensional translation – constraint or potential?” in: Bosinelli, Rosa Maria and Christine Heiss (eds) Traduzione multimediale per il cinema, la televisione e la scena. Bologna: Clueb, 29-46

Taylor, Christopher (2003). “Multimodal Transcription in the Analysis, Translation and Subtitling of Italian Films”. In: The Translator, 9 (2), 191-205.

Taylor, Christopher (2004). “Subtitling, Filmese and MCA” in: Carol Taylor, Maria Grazia Busà and Sara Gesuato (eds.), Proceedings of the Seminar, Linguatel: Progetto telematico di lingua e mediazione interlinguistica, 9-10 May 2002, University of Padua, Padova: Unipress, 3-18

Teubert, Wofgang (1996). “Comparable or parallel corpora?”, in: International journal of lexicography 9 (3), 238-264

Ulrych, Margherita (2001). “What corpora for what pur pose: a translation-based view” in: Carol Taylor Torsello, Giuseppe Brunetti and Nicoletta Penello (eds.) Corpora testuali per ricerca, traduzione e apprendimento linguistico. Studi Linguistici Applicati. Padova: Unipress, 361-371

Valentini, Cristina (2006). “A multimedia database for the training of audiovisual translators”, in: Jostrans, 6 [online] (Last Access: 31 October 2007)

Valentini, Cristina (2007). “Developing AVT corpora for a quantitative approach to language transfer in cinematic products”, in: Mus, Federico (ed.) Selected Papers of the CETRA Research Seminar in Translation Studies 2006 [online] (Last Access: 16 June 2008)

Valentini, Cristina (2008). “Forlixt 1: The Forlì Corpus of Screen Translation: Exploring Macrostructures”, in: Chiaro Delia, Heiss Christine and Bucaria, Chiara (eds.) Between Text and Image: Updating Research in Screen Translation. Amsterdam/Philadelphia: John Benjamins, 37-50


Festa di Laurea (1985, Pupi Avati) [Graduation Party]

Pane e Tulipani (2000, Silvio Soldini) [Bread and Tulips]

La vita è bella (1997, Roberto Benigni) [Life is beautiful]

Zuckerbaby (1985, Percy Adlon) [Sugarbaby]

Mimì metallurgico ferito nell’onore (1971, Lina Wertmüller) [The Seduction of Mimì]

Nordrand (1999, Barbara Albert) [Northern skirts]


[1]The authors jointly discussed and designed the contents and style of the entire paper. However Cristina Valentini is mainly responsible for section 1 to 3, and Sabrina Linardi is mainly responsible for section 4 to 5.

[2]Current members of the research group are: Christine Heiss and Marcello Soffritti, the initiators of the project; Sabrina Linardi, in charge of data entry and updating; Piero Conficoni, the software developer; and Cristina Valentini, who equally carried out data entry and updating activities.

[3]Author’s translation.


[5]To date the corpus is made up of different film genres (comedy, drama, thriller) and it benefits from a differentiation between proper cinematographic genres and TV genres (detective stories, soap-operas, etc.).

[6] Last update: October 2008.

[7]Tran(scripts) of films have been provided partly by graduating students, who chose to investigate different aspects of audiovisual translation in their final dissertations; partly by dubbing studios as far as post-production scripts are concerned and specialized magazines (e.g. L’Avant-scène cinema); partly have instead been transcribed by the authors of this article.

[8]This choice has been possible thanks to the availability of the script directly provided by the dialogist/adaptor.

[9]For further details and examples about the full-text search and the guided search (or search by attributes) see also Valentini 2006, 2007, 2008, and Heiss and Soffritti 2008.

[10]For further details see Aldeghieri Chiara 2000.

About the author(s)

Cristina Valentini got her degree in Conference Interpreting from the SSLMIT (University of Bologna), in 2001, with a dissertation on a survey conducted among simultaneous interpreters to assess the use of new technologies in the booth. In 2003, after having completed a five month traineeship at the SDT of the European Commission, she became research assistant at the SITLEC (University of Bologna at Forlì), specializing in the development of terminology and multimedia databases. In January 2006 she has been awarded a Ph.D. scholarship from the Department of Interdisciplinary Studies in Translation, Languages and Culture (SITLEC- Forlì) with a research project in audiovisual translation corpora. Her research interests currently focus on multimedia corpora, dealing particularly with polisemiotic aspects of film translation and film discourse and on technical and legal terminology pertaining to the specialised domain of health and safety at work. She has attended several international conferences to present her studies (Saarbrücken 2005, Barcelona 2005, Bergamo 2005, Forlì 2005, Copenhagen 2006, Berlin 2006, Montréal 2007, Forlì 2007).

Sabrina Linardi got her degree in Translation from the SSLMIT (University of Bologna), in 2002. 2002-2003 she attended the first master in ScreenTranslation at the SSLMIT and got her degree in Screen Translation inNovember 2003. Since 2004 she has been collaborating with the SITLEC(University of Bologna at Forlì) for the project of a multimedia database.2006 She became research assistant at the SITLEC, specializing in thedevelopment of a multimedia database. She has attended the internationalconference together with Cristina Valentini to present theirstudies on FORLIXT 1 (Multimedialectranslation, Forlì 2007).

Email: [please login or register to view author's email address]

©inTRAlinea & Cristina Valentini & Sabrina Linardi (2009).
"Forlixt 1 : A multimedia database for AVT research"
inTRAlinea Special Issue: The Translation of Dialects in Multimedia
Edited by: M. Giorgio Marrano, G. Nadiani & C. Rundle
This article can be freely reproduced under Creative Commons License.
Permanent URL:

Go to top of page