Annalisa Sandrelli’s research while affiliated with University of International Studies of Rome and other places

This paper discusses the potential use of Automatic Speech Recognition (ASR) tools to produce intralingual subtitles for broadcasting purposes. Two different ASR tools were trialled by an international broadcaster to produce automatic subtitles for pre-recorded content in English and in Italian, a British talk show and a US feature film dubbed into Italian. A study was commissioned to compare the performance of the two tools on the materials. Our evaluation focused on two key dimensions: the accuracy of the transcript and the readability of the subtitles. Accuracy was assessed quantitatively by using an adaptation of the NER and NTR models (Romero-Fresco & Martínez 2015, Romero-Fresco & Pöchhacker 2017), which focuses on ASR-generated errors and categorises them by error type (content- or form- related) and by level of severity (minor, standard and critical). Readability was assessed qualitatively by analysing text segmentation, namely line breaks and subtitle breaks. Our findings indicate that all the ASR outputs fell short of the 98% accuracy threshold expected in the broadcasting industry, although performance was notably better in English. Moreover, subtitle segmentation and timing were found to be relatively poor in the subtitles produced by both tools in both languages. Therefore, the ASR-generated subtitles from the samples provided by the broadcaster can only be considered an intermediate step. Substantial human input is required before the tools can be put to work (customisation) and after the ASR has generated the subtitles (human post-editing) to produce broadcast-ready subtitles. Lay summary This paper explores using Automatic Speech Recognition (ASR) tools to create intralingual subtitles (i.e. in the same language) for pre-recorded content. The study was commissioned by an international broadcaster to test two ASR tools for generating subtitles in English and Italian, covering a British talk show and a US film dubbed into Italian. The study compared the tools' performance, focusing on two dimensions, i.e. subtitles’ accuracy and readability. The former, i.e. accuracy, was measured using a model that enabled us to categorise and weigh errors generated by the ASR tools. The latter, i.e. readability, was measured by considering lines and subtitles breaks. The evaluation revealed that both tools fell short of the industry's expected 98% accuracy, especially in Italian. Additionally, subtitle segmentation and timing were found to be subpar in both languages. Consequently, substantial human involvement, including customisation and post-editing, is necessary to produce high-quality broadcast-ready subtitles.

Despite the popularity of legal dramas, there is little research on the popularisation strategies used in them to convey complex legal concepts to viewers. Moreover, the translation of such audiovisual material is also relatively under-researched. This paper presents a qualitative study investigating the popularisation strategies employed in the US legal drama The Good Wife and the translation strategies used in the Italian dubbed version. The English dialogues of 12 episodes were transcribed along with their matching Italian dubbed dialogue, and the relevant legal references were extracted. When lay people interact with legal professionals on screen the dialogue is aimed at both the lay characters and the audience, and it was found that several types of popularisation strategies are used. By contrast, in scenes involving just legal professionals in conversation it would not be realistic to include too much legal exposition. As regards translation, the key principle driving the Italian dubbed version is the need to create plausible-sounding dialogue rather than a legally accurate translation. The findings of this study contribute to research on the characteristics of popularisation discourse in legal dramas and have the potential to inform the training of audiovisual translators.

In recent years, cloud-based environments have radically altered traditional subtitling workflows that can now be conducted online by virtual teams of professionals working together from all corners of the Earth. Moreover, automatic speech recognition (ASR) and machine translation (MT) tools are being integrated into audiovisual translation workflows. As a result, new professional profiles are emerging, but the new opportunities offered by technological progress are accompanied by significant challenges. The present chapter illustrates the main results of ¡Sub!: Localisation Workflows that Work (2020–2021) and its follow-up ¡Sub!2 (2021–2022), two international pilot projects carried out at UNINT university (Italy). A series of experiments was conducted on the OOONA cloud subtitling platform to compare three workflows which differed in their degree of automation. 24 subjects were recruited from amongst postgraduate students and recent translation graduates from UNINT and Roehampton universities. There were three subtitling teams working from English into Italian and three teams working from Spanish into Italian, and each team comprised a Project Manager, a Spotter, a Translator, and a Reviser. The teams were required to subtitle three 10-min clips from a science documentary according to the instructions provided for each workflow. As well as submitting the target language subtitles for each clip, the teams were required to document all the steps of the process via a team logbook, screen recordings of work sessions, Quality Assessment forms, and workflow reports. In addition, pre- and post-experiment questionnaires were administered to participants. All the data thus collected has been analysed from a quantitative and qualitative point of view to determine the most efficient workflow, i.e., the one that ensures the best quality output in the tightest turnaround time. The results of the two pilot studies will be used to inform translator training practices, to ensure they are in line with constantly evolving market demands.

Over the last few years, cloud-based environments have simplified traditional localisation workflows and have made it possible for virtual teams of audiovisual translation (AVT) professionals to work together from all corners of the earth ( Díaz Cintas and Massidda 2019 ). In addition, AI-powered technologies have been integrated into localisation workflows to accelerate translation processes: this has led to a progressive automation of AVT practices and has created brand new roles for language professionals. This paper presents the preliminary results of the international pilot project ¡Sub! Localisation Workflows (th)at Work (2020–2022). A series of experiments was conducted in the spring of 2021 to compare three different workflows in the subtitling of documentaries: traditional (i.e., using only subtitling software), semi-automated (using automatic speech recognition and captioning) and fully automated (relying on automatic speech recognition, captioning and machine translation). The experiments involved twenty-four final-year MA students and recent graduates from UNINT and Roehampton University (twelve of them working from English into Italian and twelve from Spanish into Italian), in subtitling teams that included a project manager, a spotter, a subtitler, and a reviser. All the work was recorded via screencast technology and documented in a project logbook, a quality assessment form, and a workflow summary sheet. The aim of the experiments was to identify the most effective workflow equation, i.e., the one able to deliver the best quality output in the tightest turnaround time. This paper illustrates the experimental set-up and materials and discusses the preliminary results emerging from a quantitative and qualitative analysis of our data.

¡Sub!: Localisation Workflows that Work, an international pilot project involving the University of International Studies (UNINT) in Italy and Roehampton University in the UK. The aim of the project is to analyse current practices in the AVT industry and compare different workflows and related technologies in order to identify the combination of human and technological factors able to produce ‘workflows that work’. More specifically, a series of experiments comparing workflows involving different combinations of CAT, MT and automatic speech recognition (ASR) tools will be conducted to devise the most efficient workflow equation: the best quality output in the tightest turnaround time. Subjects for the experiments, collaborating on subtitling projects carried out in teams, will be recruited among postgraduate students belonging to both universities. The audiovisual materials will be selected from among educational and science docuseries and documentaries in order to test the efficiency of the tools especially on the related LSP (language for special purposes). The present project will represent the opportunity to experiment on advanced technologies, test several tools, streamline research methodology and collect preliminary data from a relatively small group of subjects. Moreover, the results of the pilot study will be used to inform translator training practices, to ensure they are in line with constantly evolving market demands.

This paper presents the key findings of the pilot phase of SMART (Shaping Multilingual Access through Respeaking Technology), a multidisciplinary international project focusing on interlingual respeaking (IRSP) for real-time speech-to-text. SMART addresses key questions around IRSP feasibility, quality and competences. The pilot project is based on experiments involving 25 postgraduate students who performed two IRSP tasks (English–Italian) after a crash course. The analysis triangulates subtitle accuracy rates with participants’ subjective ratings and retrospective self-analysis. The best performers were those with a composite skillset, including interpreting/subtitling and interpreting/subtitling/respeaking. Participants indicated multitasking, time-lag, and monitoring of the speech recognition software output as the main difficulties; together with the great variability in performance, personal traits emerged as likely to affect performance. This pilot lays the conceptual and methodological foundations for a larger project involving professionals, to address a set of urgent questions for the industry.

Focusing on the multi-faceted topic of Eurolects, this volume brings together knowledge and methodologies from various disciplines, including sociolinguistics, legal linguistics, corpus linguistics, and translation studies. The legislative varieties of eleven EU official and working languages (Dutch, English, Finnish, French, German, Greek, Italian, Latvian, Maltese, Polish, Spanish) are analyzed using corpus methodologies in order to investigate the variational dynamics and translation-induced patterns of the different languages. The underlying assumption is that, within the sociolinguistic continua of the EU languages, it is possible to single out specific legislative varieties (Eurolects) that originate at a supra-national level. This research hypothesis is strongly supported by the empirical findings derived from detailed corpus analyses of each language. This work represents the first systematic and comprehensive linguistic research conducted on a wide range of EU languages using the same protocol and applying corpus methodologies to the extensive Eurolect Observatory Multilingual Corpus.

Paolo Virzì’s film production is deeply rooted in Italian (often Tuscan) contemporary society, with a focus on the daily lives of ordinary people and the difference between their true selves and their public persona. The language spoken by his colourful characters, which includes regionalisms, slang, colloquialisms and a vast array of idiosyncratic traits, is the quintessence of Virzì’s world and of his bittersweet irony. In this contribution, we aim to illustrate how these culture-specific scenarios are transposed in the English subtitled versions of three recent Virzì films, namely La prima cosa bella [The first beautiful thing] (2010), Paolo Virzì, Italy, Tutti i santi giorni [Every blessed day] (2012), Paolo Virzì, Italy, and Il capitale umano [Human capital] (2014), Paolo Virzì, Italy. We identify the indices of regional or Italian identity, linguistically or culturally expressed, in the three aforementioned films, with a view to elucidating the difficulties behind their translation in the English subtitles.