How to Tackle a Transcript

From TED Translators Wiki
(Redirected from Transcribing talks)
Jump to: navigation, search
Read this article in other languages: Español日本語French

A TEDx transcript is a form of same-language subtitles or captions. In addition to containing the words spoken by the speaker, the transcript must additionally be divided into subtitle lines and then synchronized (timed) to match the flow of the recorded talk. Like closed captions, TEDx transcripts also contain sound information for Deaf and hard-of-hearing viewers. Below, you will find hints and strategies useful in creating TEDx transcripts as an OTP volunteer. If you haven't joined the OTP yet, go to TED.com/transcribe.

This guide is an extension of this video tutorial. Note that the line-length and reading speed information below are guidelines for languages based on the Latin script; for other languages, the rules may be different. If you believe these rules are not suitable for your language, please contact us at translate@ted.com.

IMPORTANT: before you start working on a transcript, make sure that the video is part of the TED team on Amara, using this guide (which also contains a link to a form you can use to add a video that is not on Amara). Otherwise, it may be impossible to publish your work on YouTube and make it available for translations. This tutorial shows how to properly search for talks available for transcription on Amara.

What are the benefits of getting your talks transcribed?

Transcripts are important for several reasons:

  • Same-language subtitles make the talk accessible to Deaf and hard-of-hearing viewers
  • Transcribed talks get indexed in Google, giving them and your event more exposure
  • Only talks with a transcript can later be translated (and possibly considered by TED for further distribution)

The transcription project workflow

TEDx talk videos are uploaded to YouTube. Subtitles for those videos are created in an online tool created by our subtitling partner, Amara. In order to sign up for an account on Amara, and learn how to find videos to subtitle, watch these short OTP Learning Series tutorials.

Once a transcript has been completed, it must be reviewed by another volunteer and then approved by a Language Coordinator. Approved transcripts can then be viewed while watching the TEDx talk on YouTube. The transcriber and reviewer are credited for their work on their TED.com profiles.

To get additional support, consider joining the general Facebook group for Open Translation Project volunteers, and/or the local TED translator group for your specific language. You can find the list of language groups here.

HINT: If you're working on an English transcript, make sure to read our English Style Guide.

Overview of the transcribing process

Image shows the controls box in the Amara interface.
Users can review controls and guidelines right from the subtitling interface
Transcribing an 18-minute talk usually takes between 4 to 6 hours; the user has 30 days to complete that task. Transcribing is divided into three steps:

1. Writing down text and splitting it into subtitles
This step usually takes between 2-4 hours and involves typing down what the speaker says and dividing this text into subtitles that are in keeping with TED’s standards for length and are easy to read (e.g. don’t contain slips of the tongue, don’t merge two sentences together).

2. Synchronizing the subtitles, editing the reading speed
This step usually takes up to one hour. The transcriber uses a simple interface to mark where the subtitles created in step one should display, and then fine-tunes the timing where necessary to improve synchronization and bring the reading speed down to TED’s standards.

3. Editing the title and description
Before submitting the subtitles, the transcriber needs to make sure the title and description of the talk are in the language of the talk and are formatted according to TED’s standards (learn more here).

To get a quick overview of working with subtitle lengths and reading speed, watch this short video tutorial, as well as this tutorial that contains a few useful tips for transcribing talks. Below, you will find more detailed advice covering each of the three transcribing steps, as well as some more technical information on formatting and timing the subtitles.

Below, you will find hints and strategies that you will find very useful when transcribing talks. For a quick introduction, watch this short video tutorial.

Dividing the text into subtitles

This step usually takes between 2-4 hours. The user plays the talk and types down what the speaker says. In order to allow the viewer to read the subtitles easily, while typing down the transcript, the transcriber breaks subtitles longer than 42 characters into two lines, and begins a new subtitle once a maximum of 84 characters total have been reached (the subtitle can be shorter). This length information is displayed conveniently in the subtitling interface, for every subtitle. (Note: these values are applicable to all languages that use the Latin script. For length standards in other languages, consult resources in that language’s section of OTPedia or ask a Language Coordinator).

The main goal is to create subtitles that are easily read, well-rounded bits of text. This means that transcribers try to only split subtitles where it wouldn’t separate phrases and grammatical units (e.g. they don’t split an article and a noun at the end of a line or subtitle). To comply with TED’s length and line-breaking standards, a degree of rephrasing is permissible, as long as it doesn’t change the meaning of the sentence; slips of the tongue and obvious mistakes should not be included in the transcript.

Image shows the transcribing step in the Amara interface.
The first step is about dividing text into subtitles

When deciding how to divide the text into subtitles, you should consider the following points:

1. Is the subtitle long enough to break it into two lines?
If the text you will have in the subtitle is over 42 characters in length, you should break it into a maximum of two different lines (two lines in the same subtitle). To break the line, hit Shift+Enter. You don’t need to break subtitles shorter than 42 characters; very short subtitles broken into two lines can be distracting to the viewer. IMPORTANT: The subtitle should never be longer than 84 characters total, and should contain no more than 2 lines.

2. Is the text that I'm entering too long to work as a single subtitle?
If the text you are entering is longer than 84 characters, you should create two subtitles instead.

3. Do the lines and the whole subtitle end neatly in "linguistic wholes"?
You should take care to break the lines and end the subtitles after linguistic wholes (e.g. don’t separate a possessive and a noun or somebody’s first and last name). Learn more here.

4. Am I including redundant text?
Broken phrases ("I wanted to--No, this is what I'll talk about"), repetitions ("Thank you, thank you, thank you, thank you") and empty syllables ("erm," "umm" etc.) should not be included in the transcript. Also, do not include obvious errors, like when the speaker says "We thinks" instead of "We think." Instead, use the correct form of the word in the subtitle. On rare occasions, if you believe that the need for the change is obvious (e.g. the speaker says “up” instead of “down”), but your edit will significantly alter the meaning of the sentence, put it in square brackets, to indicate intentional editing (e.g. “I woke up at 9 AM, and the sun was [up].”).

5. Do I really have to cut the sentence up into this many subtitles?
As much as possible while respecting the length and reading speed standards, try to have the subtitle contain a “full” part of the sentence (a clause), or the whole sentence. This will make it easier to read, and it will be easier for translators later on to translate bigger chunks of one sentence than smaller ones, since not everything will divide up easily in the same way in the target language as it does in the original. To learn more about how to make your subtitles easier for future translators, see this guide.
IMPORTANT: Never include the end of one sentence and the beginning of another in the same subtitle (e.g. "this is why./And another idea").

This printable cheat sheet contains all of the main OTP technical subtitling standards for Latin-script languages

6. Did I include all of the sound information essential to understanding the talk?
Include all of the sound information essential to understanding the talk, such as non-verbal sounds that the speaker refers to (“(Clears throat) Sorry about that.”), off-screen speaker changes (indicate who is speaking, if that is not obviously visible), as well instances of music, clear laughter and applause from the audience (with the exception of intro music and applause heard at the beginning of the talk). Also, indicate any temporary change of language, and translate the subtitle into the main language of the talk (e.g. “(Arabic) This is my idea.”) Put the sound information in parentheses (e.g. (Music)), with the first letter capitalized, and always represent the sound, not the event that caused it (e.g. “(Gunshot),” not "(Dog fires gun)."). For more information about using sound representation, read this guide.

7. Did I include on-screen text?
If possible without overlapping other subtitles and going over the subtitle length and reading speed limits, include on-screen text that is part of the talk (e.g. text on slides or embedded subtitles in a video played on the stage). This will allow this text to be translated into other languages. In order to signify that this is on-screen text and not something the speaker is saying, put the representation of on-screen text between square brackets.

Do not transcribe on-screen text which is not relevant to the content of the talk, nor text which will not be translated (e.g. the name of the TEDx event).

Synchronizing the subtitles with the video

Image shows the synchronization step in the Amara interface.
The subtitles are synced using a simple videogame-like interface
This step usually takes up to one hour. Starting with text neatly divided into subtitles, the transcriber now needs to tell the system when to show each of the subtitles while playing the video. The user plays the talk and hits the up arrow when the first subtitle should start displaying, and then hits the down arrow whenever the currently highlighted subtitle should stop displaying and the next one should start. Afterwards, they go back and make finer edits to the timing using sliders on the video timeline to set the beginning and end of subtitles (e.g. to fix a subtitle that starts displaying too long after a speaker started the equivalent sentence). For more information on using the Amara interface to sync subtitles, read this article.

Once the subtitles have been synchronized, the user goes back to implement reading speed fixes using sliders in the timeline. In order to allow the viewer to read the subtitle while it’s displayed on the screen, the reading speed for each subtitle must not be higher than 21 characters per second. This speed information is displayed for every subtitle on Amara, and wherever this speed is exceeded, the transcriber can compress or reduce text (without changing the meaning) or/and extend the duration of the subtitle to fix the issue.

HINT: A red exclamation mark is displayed on every subtitle that needs fixing for length or reading speed.

When synchronizing your subtitles, consider the following points:

1. Is the reading speed no more than 21 characters/second?
The maximum reading speed for subtitles is 21 characters/second. To maintain a good reading speed, you can extend the duration of the subtitle, even if it’s going to run a little into the time the next sentence is spoken (but don't start the subtitle more than about 100 ms before the equivalent bit of speech is heard).

Extending the duration usually helps, but if necessary for a good reading speed, combine this with rephrasing the text of the subtitle to shorten/compress it while preserving the meaning. Remember that with a reading speed that is too high, the subtitle will just disappear too quickly for most viewers to read, which is tantamount to cutting it out of the transcript. For this reason, it’s always better to compress the text a little rather than create a verbatim transcript that viewers won’t be able to follow. Good reading speed is also very important because your transcript will often serve as the starting point for translations, and the equivalent subtitle may become much longer in the target language, raising the reading speed even more.

For more advice on compressing/reducing text in subtitles, see this guide.

HINT: Occasionally, if the subtitle contains potentially difficult vocabulary (scientific terminology, obscure proper names), consider lowering the reading speed to values even below 21 characters/second, to make it easier for the viewer to take in the content of the subtitle and allow more reading speed for future translations (which are often longer than the original subtitle).

2. Is the subtitle synchronized with the equivalent bit of speech?
Generally, the subtitle should start displaying when the speaker says the equivalent bit of speech. However, good reading speeds are more important than perfect synchronization. If you need to extend the duration of the previous subtitle to get a good reading speed, it’s OK to have the next one start some time after the speaker said those words. However, don’t have the subtitle start displaying before the speaker says the equivalent sentence, since the mismatch in body language and on-screen content can be distracting to the viewer. This is especially important in cases where synchronizing changes in the video with changes in the subtitles is crucial to what happens in the talk (e.g. if possible, a subtitle that reveals what's in a slide should not show up before the slide shows up on the screen).

3. Is the subtitle’s duration shorter than 1 second or longer than 7 seconds?
A subtitle displaying for less than one second will usually disappear too quickly for most users, and this issue will be compounded in translation. Subtitles displaying for over 7 seconds are distracting to the viewer and should be split into two separate subtitles.

If there is a longer piece of music or applause, have the sound representation (e.g. (Music)) display for 3 seconds and then indicate when the sound is about to end (e.g. (Music ends)).

4. Does the subtitle lag too long into a pause?
Do not have the subtitle stay on the screen for more than 1 second after the speaker has paused after a sentence. If you’ve covered up long pauses in the synchronizing step, once you’re done synchronizing the whole transcript, you can shorten the durations of these subtitles using the sliders in the timeline, so that they don’t lag over pauses. You can choose not to show pauses inside a sentence, or if necessary, indicate that the sentence was broken off by using dots (...) or a dash (-), depending on the conventions in your language (note: in subtitles, use a minus instead of a full dash). However, always try to show longer pauses between complete sentences.

Avoiding character display errors: simple quotes, apostrophes and dashes

Using smart/curly double quotes (“”) is precarious, because some players will have trouble displaying them correctly. Please use the simple, straight ASCII double quote (") or the straight apostrophe ('') for single quotes. The rule is similar for apostrophes: use the straight apostrophe (') instead of the typographic/curly apostrophe (’). Instead of an en/em dash (–/—), use a hyphen (-).

For other punctuation marks in your languages, as much as possible, use a simple ASCII equivalent (research to find one for your language). This may go against strict typographic conventions, but the technical limitations of most subtitle formats mean that without this simplification, for some users, many of the "correct" characters will simply not be displayed (e.g. when playing talks offline). Note that these rules only apply to the subtitles, and you should use proper punctuation in titles and descriptions.

You should not use HTML tags or any other formatting tags in TEDx transcripts, because these tags will not display correctly in the YouTube player.

Title and description format

Each TEDx talk comes with a title and description added by the TEDx organizer, which are imported into Amara from YouTube. However, these sometimes contain too little or too much information and may not conform to the formatting standards described below. In these cases, you are expected to edit them before you submit your transcription.

Note: The language of the title and description should match the language of the talk. Do not put English titles and descriptions on non-English talks.

Title format

Image shows how to edit the title and description of the talk in the Amara interface.
Click the “pencil” button to edit the title and description

The standard title format uses the talk’s title, the speaker’s name and the TEDx event’s name, separated with the vertical bar (pipe) character (with a space before and after it):

On being a young entrepreneur | Christophe Van Doninck | TEDxFlanders

If the title is formatted differently, modify it to match the standard format. Do not add the event’s date to the title.

If the title is missing, it's OK to just leave the speaker's name, but consider coming up with a title on your own or contacting the organizer or speaker for a title suggestion.

In English titles, use sentence case: capitalize only the first word in the title and any proper names.

Description format

The description should consist of a short overview of the talk. Remove all links to external websites (unless they represent the speaker’s organization that the talk is about). If the description also contains the speaker’s bio, you can keep it in, but the general text explaining what the TEDx program is should be left out (“In the spirit of ideas worth spreading, TEDx is a program of local, self-organized events…”). If the description is missing, please consider adding your own short description of the talk.

The description may also contain the following disclaimers, which should be kept in and translated:

This talk was given at a local TEDx event, produced independently of the TED Conferences.

This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at http://ted.com/tedx

Here, you can find model translations of these disclaimers in various languages. If you can't find your language, consult with a Language Coordinator and send the model translation that you came up with to translate@ted.com.

How to get more talks transcribed

If you are a TEDx organizer with multiple untranscribed talks, consider reaching for help out to the TED Translators and Transcribers community on Facebook. Try to select one or two prioritized talks and explain why it’s important for you to get these particular talks transcribed. Find ways to make transcribing your talks a challenge and make sure to show appreciation to the transcribers (e.g. by thanking them on your website).

Remember that the volunteer TED transcribers and translators are volunteers and they usually select talks that are meaningful to them in some way, out of the tens of thousands of TEDx talks in the world. Because your team and your local community are much more invested in trying to promote the ideas in the talks from the events they have attended, try to collaborate with the local transcriber community in coaching your team in transcribing talks and organizing transcribeathons.