Project Website
Duration
2017-2019
Principal Investigator
Elena Davitti (University of Surrey, GALMA)
Participants
Annalisa Sandrelli (UNINT, GALMA)
Pablo Romero-Fresco (UVigo, GALMA)
Funding body
University of Surrey
Budget
€25,000
Description:
Intralingual respeaking (a technique for live subtitling whereby respeakers listen to live input and simultaneously repeat it the same language to a speech recognition software that turns it into written subtitles) is now widely used to provide subtitling for the deaf and hearing-impaired in many settings, especially on television. By contrast, interlingual respeaking is still in its infancy. As respeaking between two languages has the potential to address the pressing need to access an ever-increasing amount of multilingual content while providing a fundamental service not only to deaf communities but to linguistic and cultural communities and the general public at large, much research is required on (1) the skills and competences needed by interlingual respeakers and on (2) how to assess the quality of their output. The key question is whether a specific training background can support the acquisition of the skills needed to perform this challenging task. The similarities and differences between simultaneous interpreting (SI) and respeaking have been explored in a number of experimental studies, but very few contributions are focused specifically on interlingual respeaking. This project aims to contribute to this growing body of research by comparing the interlingual respeaking performance of three groups of trainees, recruited at the University of Surrey and at UNINT:
- 1. with training in both SI and (intralingual) respeaking
- 1a. advanced level (graduate – no3)
- 1b. intermediate level (half-way through course – no3)
- 1a. advanced level (graduate – no3)
- 2. with training in SI but not in respeaking
- 2a. advanced level (graduate – no3)
- 2b. intermediate level (half-way through course – no3)
3. with an interpreting background excluding SI and no training in respeaking (control group – no3)
- 2b. intermediate level (half-way through course – no3)
The experiment took place at the beginning of 2018 and consisted of the same set of interlingual exercises for each profile, performed using the Dragon Naturally Speaking software. Subjects without a respeaking background were provided with an introduction to the software to enable them to create a user profile and operate its basic functions. Output quality was measured with the newly developed NTR model for quality assessment in interlingual RSP. Scores were correlated with users’ subjective ratings through qualitative methods to investigate whether SI skills support the task. This pilot experiment is part of a wider project currently being planned and its results are expected to provide useful indications to inform further, larger-scale research activities involving trainees and professionals.
Links:
Presentation at UMAQ conference: https://drive.google.com/open?id=1zgfyirr7hUDTUCxynNyR8C7bEzOPmeKG