Respeaking is commonly used to subtitle live TV programmes for the benefit of deaf and hard-of-hearing viewers (SDH subtitles): it is based on the interaction between the respeaker's competences and a speech recognition software trained to recognise his/her voice. In some countries this technique is also used to subtitle live events, such as conferences, debates, and so on. In recent years there have been attempts to use the interlingual variant of the technique, thus making it possible to translate a speech from one language to another (as you do in simultaneous interpreting) and obtain real-time subtitles. The respeaker listens to the incoming message and interprets it simultaneously to the software, adding punctuation verbally by means of voice commands; then, the respeaker's audio input is processed by the software, which produces the transcript; the respeakerchecks the output and corrects any errors, and the subtitles are finally made available to the audience. After a short overview of the experimental research carried out in this field and the available theoretical models, the present paper assesses the accuracy of the interlingual subtitles (English into Italian) produced via IRSP in a live event (a symposium). The analysis outlines the specific difficulties of this technique when applied to a live event, with the aim of producing useful data and suggestions for the development of practice guidelines.
Il respeaking è usato comunemente in televisione per sottotitolare i programmi in diretta per il pubblico sordo e audioleso: si basa sull’interazione tra le competenze del respeaker e un software di riconoscimento del parlato addestrato a riconoscerne la voce. Questa tecnica viene usata in alcuni paesi anche per sottotitolare alcuni eventi dal vivo, come conferenze, dibattiti e così via. Di recente si è iniziato a sperimentare anche il respeaking interlinguistico (IRSP), che consente di tradurre un discorso orale (come nell’interpretazione simultanea) ottenendo una trascrizione in tempo reale: il respeaker ascolta il messaggio e lo traduce simultaneamente per il software, aggiungendo anche la punteggiatura mediante dei comandi vocali; dopodiché, l’input vocale del respeaker viene elaborato dal software che produce la trascrizione; dopo aver controllato e corretto eventuali errori, i sottotitoli vengono trasmessi al pubblico. Dopo una breve panoramica delle ricerche sperimentali esistenti in tema e dei modelli teorici di riferimento, il presente studio presenta l’analisi dell’accuratezza dei sottotitoli interlinguistici (inglese-italiano) prodotti mediante IRSP in un evento dal vivo (un simposio), applicando per la prima volta il modello NTR a dati reali. L’analisi delinea le difficoltà specifiche di questa tecnica utilizzata in un evento dal vivo, con l’obiettivo di fornire dati utili e indicazioni per lo sviluppo di linee guida in merito.
Eventi dal vivo e accessibilità: uno studio di caso sul respeaking interlinguistico
SANDRELLI A
2021-01-01
Abstract
Respeaking is commonly used to subtitle live TV programmes for the benefit of deaf and hard-of-hearing viewers (SDH subtitles): it is based on the interaction between the respeaker's competences and a speech recognition software trained to recognise his/her voice. In some countries this technique is also used to subtitle live events, such as conferences, debates, and so on. In recent years there have been attempts to use the interlingual variant of the technique, thus making it possible to translate a speech from one language to another (as you do in simultaneous interpreting) and obtain real-time subtitles. The respeaker listens to the incoming message and interprets it simultaneously to the software, adding punctuation verbally by means of voice commands; then, the respeaker's audio input is processed by the software, which produces the transcript; the respeakerchecks the output and corrects any errors, and the subtitles are finally made available to the audience. After a short overview of the experimental research carried out in this field and the available theoretical models, the present paper assesses the accuracy of the interlingual subtitles (English into Italian) produced via IRSP in a live event (a symposium). The analysis outlines the specific difficulties of this technique when applied to a live event, with the aim of producing useful data and suggestions for the development of practice guidelines.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
