Lingsoft has been producing subtitling in Finnish and Swedish since 2011. I have been part of the subtitling team since the beginning. Subtitling means monolingual closed captioning – for example, providing Finnish subtitles for programmes in Finnish and Swedish subtitles for programmes in Swedish. Subtitling is meant for all those who, for some reason, cannot hear the speech in the programme or wish to watch it without sound.
Why do we need subtitling?
Subtitling is an important resource for achieving equality and accessibility in a society. Without subtitling, some people don’t have equal access to current events and phenomena.
The internet is full of all kinds of videos. What if you left your earphones at home? When a video has been subtitled, you can watch it during a bus or train ride or in an open office without disturbing those around you.
What makes good subtitling?
Good subtitling doesn’t just mean writing down everything you hear. Subtitling would be very hard to follow if everything were written down exactly as you hear it. Careful summarising and editing sentence structures are important parts of the process. Good subtitling is fairly close to the actual speech being subtitled but the text should be easily understandable and well written.
The content of the programme and its intended audience are of course significant in determining, for example, if spoken language expressions are to be used in the text.
If the viewer cannot hear the speech in the programme, they won’t know when the speaker changes. This must also be taken into account in subtitling to make sure that the viewer knows what’s happening.
The viewer must also be told if there are other sounds besides speech that may be important for the programme. For example, laughter from the audience, the sound of a gunshot, or the ringing of a phone are usually important information.
At its best, high-quality subtitling is created in cooperation between people and machines
In subtitling, Lingsoft uses its own speech recognition software, which creates a preliminary version of the subtitles. This preliminary subtitling includes a transcription of the speech as well as preliminary timing and division for lines. This allows the human subtitler to focus on editing the text to make it more compact and understandable and to improve the quality of the language and timing to meet the quality requirements set for subtitling. Automatic speech recognition works best in scripted programmes, such as children’s programmes or narrated documentaries. In such a case, it can save a lot of time.
We at Lingsoft pay a lot of attention to the quality of subtitling, and we have participated in a national work group to create unified quality guidelines for subtitling. The guidelines can be found here (currently available in Finnish and Swedish).
Subtitling Service Coordinator, Lingsoft