Accuracy Rate in Live Subtitling: The NER Model

Abstract

Over the past few years, the focus of audiovisual translation (AVT) seems to have shifted from quantity to quality. As is demonstrated by international conferences, such as Media for All 3 in 2009 ( http://www.mediaforall.eu /all3) and Media for All 4 in 2011 ( http://www.imperial.ac.uk /humanities/translationgroup/mediaforall4), this shift applies to industry as well as to academia. In the case of live subtitling, and more specifically respeaking, the most common method used to evaluate the quality of subtitles produced in real time consists of assessing their accuracy. Needless to say, where quality is concerned there are also a number of other features to be considered, such as delay, positioning, character identification and speed, as well as factors relating to their reception by viewers (opinion, comprehension, perception). These issues have all been discussed by Romero-Fresco (2011) with particular reference to the UK market. Yet, what concerns broadcasters, regulators such as Ofcom and subtitling companies is the accuracy of live subtitles and it is this that constitutes the main focus of this chapter. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 98 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.