3. Live-Subtitling: Respeaking and Speech-to-Text Interpreting via Keyboard

Summary

In the field of Live-Subtitling, two key methods are used to assist people with hearing impairments who have no or very little knowledge of sign language: Respeaking (Speech-to-Text Interpreting with speech recognition) and traditional Speech-to-Text Interpreting, in which spoken text is entered via a keyboard to produce written language. Live-Subtitling is applied in a number of different In the field of Live-Subtitling, two key methods are used to assist people with hearing impairments who have no or very little knowledge of sign language: Respeaking (Speech-to-Text Interpreting with speech recognition) and traditional Speech-to-Text Interpreting, in which spoken text is entered via a keyboard to produce written language. Live-Subtitling is applied in a number of different settings, from public talks to debates, from live television broadcasts to university lectures and school classes. In most of these cases, it is hardly possible to anticipate what will be said by the speaker, if at all. Thus, it is not possible to resort exclusively to traditional, pre-produced subtitles.

In Respeaking, the speaker’s original utterances are reformulated by the respeaker and converted into written text by speech recognition software. Currently, only speaker-dependent speech-to-text systems provide adequate results. This means that the system must first be trained to fine-tune the recognition of the respeaker’s voice, resulting in increased accuracy.

Since the very act of Respeaking can cause disturbance in the classroom, traditional Speech-to-Text Interpreting is often preferred in teaching. The lecturer’s spoken text may then be displayed as written text on a presentation slide, on an extra screen (for multiple hearing impaired people), or directly on the student’s laptop monitor.

In both forms of Live-Subtitling, time pressure and the process of transferring spoken language into punctuated written language make it necessary to summarise and condense the source text. During this process, the original information content should be altered as little as possible (Jekat et al. 2015).

Our practice partners in this Research Area include the Respeaking Department of Swiss Television SRF, SWISS TXT AG, pro audito schweiz and other organizations for hearing-impaired people.

The main objectives of Research Area 3 are therefore:

• To compare the two Live-Subtitling methods (i.e. Respeaking and traditional Speech-to-Text Interpreting) and clarify their possible uses in higher education settings.
• To evaluate and build on existing quality assessment models for live subtitles (e.g. in broadcasting corporations). Research in this area will also inform the revision of the ZHAW’s teaching modules in Live-Subtitling and the training course on traditional Speech-to-Text Interpreting developed by the ZHAW in cooperation with pro audito schweiz.
• To develop recommendations for a) standardized live-subtitling services in Swiss German, and b) the application of Respeaking in French- and Italian-speaking Switzerland. The translation of spoken Swiss German dialects into written Standard German is a particular challenge: in this case, the translation from the spoken to the written code, which is intralingual in principle, shares many similarities with interlingual simultaneous interpreting.

 References

• Jekat, Susanne J., Heike E. Jüngst, Klaus Schubert, and Claudia Villiger, eds. 2015. Themenheft Barrierefreie Sprache in der digitalen Kommunikation für Öffentlichkeit, Institutionen und Unternehmen. trans-kom 8, no. 1. http://www.trans-kom.eu/ihv_08_01_2015.html. Last access: 28.12.2017.