Transcription Guide (General)

Update: May 2021

 2

Welcome aboard! 3

Manual Transcriptions 3
1.2 Billing & Payment 4

2. The workflow summarized 5

3. The general rules of transcription 6

3.1 The general rules of transcription 6

the differences 6
Clean-read 6

Verbatim 7

3.2 The specific features of transcription 8

General transcription rules 9

Punctuation 9

The digital 9

The time format and additional indications 10

4. LaPerfect Platform & Amberscript Editor 11

4.1 The dashboard of the transcribers 11

Current jobs 11

Job invitations 11

Accepted jobs 12

4.2 Working with AmberscriptEditor 12

Identification of speakers 12

Timestamp 13

The Highlight function 13

Keyboard shortcuts 13

4.3 Submission & quality control 13

5. Tips and tricks 14

6. Problem solving 15

6.1 First, look for the answer yourself on our shared drive. 15

6.2 The main problems of transcription jobs 15

Audio, what to do in case of... 15

Editor, what to do in case of... 16

 3

Welcome aboard!

Amberscript builds SaaS solutions that allow users to

automatically transform audio and video files into text and subtitles
through voice recognition. We use the data that our users
generated to improve the best voice recognition software of
European languages. With our online editor and our human transcribers,
we are reaching transcriptions with a 100% accuracy.

Every day, we work hard to get closer to our goal: to make audio
accessible. We are absolutely convinced that a profit-oriented goal can go from
pair with the goal of creating a great positive impact on the world.

Our solution is a blend of three disciplines: developers who implement

work of innovative technologies, linguists who decode complexities
of language and scientists who discover the latest advances of
Artificial Intelligence.

Our voice recognition software is quite advanced, but is not (yet)

perfect! To be able to make perfect transcripts and subtitles,
Amberscript uses an optimal combination of A.I. technology and humans.
With a team of over 500 native speaker experts, we deliver
manual transcriptions and translated subtitles to our clients.

1.1 Manual Transcriptions

Amberscript has over 500 professional native transcribers who transform audio.
and the video into perfect text, assisted by our voice recognition software. Our
Transcribers use our editor to make this process as efficient as possible.
possible.

● Combination of AI & human to achieve 100% accuracy

● The transcript contains the speaker identification and the timestamp.

● The transcription is 'clean read' meaning in refined reading or 'verbatim'

that is to say the exact text.

● Delivered in 3 working days

● Exportable in Word, JSON, Text format, and much more.

 4

1.2 Billing & payment

At Amberscript, our transcribers are paid for every minute of audio.

You are responsible for invoicing us for the work you carry out.
Therefore, it is important to keep track of the jobs you have.
performed, as well as the duration and type (for example 'clean-read' or 'verbatim').

Invoicing explained
1. Make sure to record your completed jobs, the duration, and the type included (for example, 'clean')
read - clean reading" or "verbatim - exact text".
2. Fill in the completed jobs on an invoice (we have created a templatehere)
3. Send the invoice (in PDF) toinvoices@[Link](you always receive
a confirmation to confirm the receipt
4. You will receive your payment in 14 days by bank transfer or PayPal.
(depending on the chosen payment method on the invoice)

In most cases, you are required to pay taxes on your salary. Please
follow the instructions below to find the best solution for you:

You are not employed by Amberscript and therefore you are responsible for
pay the appropriate income taxes in your country of registration in
function of local legislation and regulation.
 5

2. The workflow in summary

The following graph shows the six-step process:

The process in brief:

1. A client downloads the audio/video for manual transcription. The client

indicate whether this should be done in 'clean read' or 'verbatim', the language of which it
it is about and the number of speakers.

2. As transcribers, you receive notifications for new jobs.

by email and/or Slack. All the jobs to be obtained appear in your
transcriber dashboard.

You will have 48 hours to complete the transcription work in.

AmberscriptEditor. In other words, you control and correct the text that is
automatically generated by our voice recognition.

4. Once you have finished, you submit the work for inspection.

5. One of our Quality Controllers will verify your work. The Quality Controller will do it.
in the following manner:

6A. If he does not accept the quality. You will be informed by email and you
You will have 24 hours to correct and resubmit the work. The
The comment from the Quality Controller is in the job Notes table.

[Link] he accepts the quality. You will also be informed by email and the
The document will be sent to the client.
 6

3. The general transcription rules

3.1 What are the general rules of transcription?

The general rules of transcription come in various ways and forms, and
may vary depending on the transcription service. The rules are therefore used for
in order to ensure that all transcriptions adhere to the same standards,
which correspond to 99% accuracy.

In this document, you will find the transcription rules of Amberscript.

3.2 the differences

Clean read (which is the most used): Objective: to capture what has been said. Final result: a
grammatically correct transcription, with a smooth reading pace.

Examples: business or organization meetings, council sessions

municipal, presentations, etc.

Objective: to capture how things are said, the way they are
Say. Final result: an accurate transcription, as close as possible to the audio.
original.

Examples: Research interviews as conversations between patients and doctors,

discourse analyses, legal documents such as hearings, interviews
of groups où the emotions of the interviewed person play an important role, etc.

Clean read
The transcribed text does NOT include:

▪ The errors in the speech

▪ The false starts (unless there is information about it)
▪ Stuttering
▪ The [Link]: do not delete the recurring repetitions of words that
I am very very happy.
▪ Interjections and filler words: these words are often used to
excessively by the speaker, but when you remove them, the sentences are
always perfectly understandable: um, ah, oops, *you know, *you see, *to be honest,
*in fact, *then, etc.

Pay attention to the context. Some of these words do not always have a function.
of interjection.

▪ The expressions must be kept regardless of the type of transcription: Oh

mon Dieu, Oh mince, Ça alors, Oh la la, oh, etc.
 7

▪ The slang words like 'Ouai', 'mouai', and 'flic' are replaced by 'Oui' or
Of course, and police. And you must correct the grammatical errors.
exemple “il y a plus de riz” devient “il n’y a plus de riz”. Ou bien encore, "il sait pas”
becomes "he does not know."
▪ Avoid starting sentences with conjunctions.
If you need to add a coordinating conjunction, extend the sentence.
For example: "I went outside but I forgot my umbrella."

NB: Regarding the clean read: Avoid all superfluous reactions like 'Yes', 'Of course.'
to maintain a fluid text, unless these reactions are responses to
questions asked.

DO NOT remove the interjections if they change the meaning of the sentence.

Verbatim
The text is transcribed exactly as it is said and includes EVERYTHING that
the speakers pronounce.

These are for example:

▪ The errors in the speech: “I went to the bank on Thursday-- no, Friday.”

▪ False starts: I, [uh], wanted-- I dreamed that I became a musician.

▪ The interjections and filler words: um, ah, oops, so, in fact, you know, you
see, etc.

▪ Slang words and grammatical errors: Yeah, chatter, cop, 'I don’t know', etc.

▪ Stuttering: I-I went to the bank last th-Thursday.

▪ The repetitions: I went- I went to the bank last Friday.

▪ Use these forms only for affirmations/negations:

Mm-hmm, Mm (affirmation) or Mm-mm (negation)
Uh-huh

Example of Clean Read:

Speaker 1: Hi, Amber, I wanted to ask you something.

Amber: Yes, of course.

Speaker 1: Someone told me that transcribers now have to pass a test.

theoretical before being able to start working. I am not sure that this
Is it true? Is that the case?

Yes.
 8

Example of Verbatim:

Speaker 1: Hi Amber, I wanted to ask you something.

Amber: Okay.

Speaker 1: Someone told me that the transcribers now have to complete a

theoretical test before being able to start working.

Amber: Yes.

Speaker 1: But I'm not sure that this is true.

Yeah.

Speaker 1: Okay, [uh], is that true?

Yeah.
 9

3.2 The specifics of transcription

The following things must be transcribed if they are relevant to the
transcription or the context.

Full Clean Read

Verbatim

The sounds that are not part of speech such as Mm-hmm Yes Yes
Mm-hmm (affirmative)

The audio file remains impossible to understand after [inaudible] Yes Yes
three listens at a high volume and a reduced speed 00:00:00- 00:00:00]
(0.7) .

Interjections and filler words like register Yes No

[uh], [ah], [oops], [so] or [in fact]. the interjection in
question

The music is played loudly and clearly - so relevant [music] Yes No

for context.

A video is launched. [video] Yes No

A continuous laugh for several seconds. [laughter] Yes No

A continuous cough for several seconds. cough Yes No

A long period of silence (3 seconds or more). [silence Yes No

00:00:00- 00:00:00]

If relevant for the transcription/context weeping Yes No

If relevant for the transcription/context [cry] Yes No

The audience applauds the speaker. applause Yes No

When several people talk to each other at the same time diaphony Yes Yes
time and thus all the speakers become
inaudible.

When the audio includes parts that are not in language Yes Yes
French. foreign
 10

3.3 General transcription guidelines.

These instructions must be followed for clean read transcriptions
verbatim.

Punctuation
. The points Each sentence ends with a period.

? The points Do questions end with an exclamation mark?

of interrogation

! The points Use an exclamation mark to emphasize a particular point!

exclamation

, The commas Commas are used for small pauses, lists, and for
improve the reading flow.

...The ellipses If there is a pause... use an ellipsis to indicate it.

- Simple tire Use a hyphen to indicate interruptions in a word

and for compound words with a hyphen like the word
robin

--Double dash Use a double dash to indicate an incomplete sentence and--

"The quotation marks" I am a punctuation enthusiast!

The digital

● The numbers zero to nine must be written in full: one, three, nine
● The number 10 and the following can be written in digits: 10, 11, 12, etc.
● If a name is at the beginning of a new sentence, it must be written in full.
Twenty-two people, Three birds, etc.
● Don't forget that the number thousand is always invariable unlike twenty and
hundreds that agree in the plural when they are multiplied by a number and not followed
Eighty apples, eighty-three apples.
● Write in digits if the number has five digits or more: 50,000
● Numbers beyond one million or one billion are written: 45 million.
● Use the terms first, second, third, etc. But for larger numbers
high use 12th, 20th, 90th, etc.
● Twenty-one, Thirty-three.
● If a price is mentioned and it is a rounded number, do not use a
€15.
● The functions and their mathematical signs are written in full: plus,
minus, multiplied, divided by, etc.
● Fractional powers are written in digits: 5 ʀ ⁄ʁ
● Use a non-breaking space for large numbers: 1 003 298 dollars
 11

● Use a comma for decimal numbers: 0.42 cents

● Phone numbers are written in digits: 06 25 25 30 30

For more details, check out this link!

Invalid input, please provide the text to be translated.
 12

The time format

● The time of the day:

○ Use numbers when the exact time is specified: 9:30
minutes.
○ Write the round hours in full: ten o'clock.
○ Avoid using AM/PM in French, and prefer 10 o'clock / 22 o'clock for
indicate whether it is morning or afternoon.
Examples:

● If the speaker says 'nine' and it implies that it is a time, you can write it down.9h / 9
hours.
● If the speaker says 'nine o'clock', writenine o'clock
● If the speaker says 'nine thirty', write it down.9 hours 30 minutes / 09:30
● Write in full the words or phrases that do not include a duration or a
horaire exactes:et demi, et quart, minuit, midi
● Use numbers to indicate degrees or radial quantities: 270
degrees, angle of 45 degrees
● infinity
degrees, phi, theta, etc.
● Timings and dates are written in numbers: 30/06/1930 or June 30, 1934

Additional indications:
Thank you for paying very close attention to punctuation, the transcription must be perfect.
punctuation marks!

Thank you for using the Editorafin spell checker to see if you might not have
forgot mistakes, it's called Toggle spell checker in the Editor.

The basic rules must be strictly followed (no errors of this type allowed)
will be tolerated!) :

● Put a capital letter after a period. Put periods at the end of each sentence.
● Accents must be respected.
● Be careful about the context: He went on vacation to Lille. / He is on vacation in
an island.
● Do not write to him if he is not available. If he is feeling fine, he will come tomorrow.

● Don't forget to name the speakers! Don't leave 'Speaker 1' or 'Speaker 2'.
but include the names of each of the speakers if you know them. If in
a file, the president is always referred to as "President" in the transcript, so
the correct notation for the speaker is therefore: "President:". A masculine voice is
indicated by: 'Mr. Name:', a female voice is indicated by: 'Mrs. Name:'. To
cases where the last names are not known, the participants are
differentiated using Mr./Mrs. + a number.
● Think about the ç! Do not write but this.
 13

4. LaPerfect Platform & AmberscriptEditor

4.1 The transcribers' dashboard

Current jobs
This tab indicates which job you are currently working on.
When you select the job file name, a number of
Information will appear such as the number of speakers, the style of
transcription, the comments from the tab, etc. From this tab, you can
go to yourEditor to edit the text or you can submit the job for
a verification when it is completed.

Note:You can only edit one job at a time. If for a specific reason you
You need to start another job before you can complete your job.
currently, please contact the operations department or explain it in the tab
Notes before submitting it for verification. Once your job is submitted, you
You can start editing a new job from your Accepted Jobs tab.

Job invitations
You will receive job invitations by email. Our way
first come, first served. The usual deadline is
48 hours.

If you receive an email but do not see an invitation, it means that

Someone faster took the job.

Thank you for always checking the Notes section for any information.
complementary.
 14

Special projects or rush files (the files with a deadline of 24 hours) are the
more often communicated on the Slack platform. So please check the channel.
specific communication in your language regularly.

In some cases, jobs can be assigned to you directly - this will not be done
only with the prior agreement of the Perfect Planners. The job can then appear in
the Accepted Jobs tab.

Accepted Jobs
In this tab, you will find all your Accepted jobs (current jobs), which refers to
your dashboard.

Currently, it is only possible to work on one file at a time, all the others
files then placed "in queue".

The usual deadline is 48 hours - we recommend starting to

work on your transcription as quickly as possible as soon as you
you accepted it.

4.2 Working with the Amberscript Editor

As our editor uses the cloud, your work will be automatically saved.
every 2-3 seconds. You can also press the Save button.
(save) on the left side of your screen.
 15

Identification of the speaker

Our recognition software vocal is capable to identify the
interlocutors/speakers. They will be by default named Speaker 1, 2, and so on.
By clicking on the speaker you can change the name. In case the name of
The speaker is known, please change the name of the speaker accordingly.
A masculine voice is indicated by: "Mr. Name:" and a feminine voice is indicated by:
Mrs. Name:

In cases where the name is not known, the participants are differentiated using
Mr./Mrs. + a number.

Timestamp
You cannot and should not manually adjust the timestamp in the Editor.
You can press the timestamp realignment button at the top.
left of your screen. This will automatically update the timestamp.

The Highlight function

In the case where in certain parts of an audio, the speaker says a word or a phrase
difficult to understand, but you think that someone more experienced
Could you transcribe it, use the highlight function and underline the word or phrase.
This will allow the Quality Controller and the client to be able to
take a look at these specific parts of the transcript.

NOTE:Remember to search for the word to avoid overusing this function!

Keyboard shortcuts
● Play and pause in your audio file: Tab or Control + Space
● Go back 5 seconds in your audio: Control + R
● Jump into the audio by clicking: Alt + Click
● Highlight the text: Control + H
● Cancel: Control + Z
● Restart Control + Shift + Z
 16

● Enable the spell checker: Control + G

4.3 Submission and quality control.

Once you have finished carefully reviewing your transcription, you can
to submit it. Each file is checked by a Quality Controller (controller
quality), which will verify your work based on the following elements:

- The accuracy of the transcription: depending on the type of transcription and its
indications.
- Linguistic accuracy: grammar, spelling, punctuation, etc.
- The search terms and specific terminologies according to the fields.

The quality of your transcription will then be accepted or rejected. If it is rejected, you
you will receive feedback, and you will have 24 hours to correct it and submit it to
new.

We work with ratings ranging from 1 to 5, with 5 being the best score.

If most of the grades you get with your transcripts are below the
average, and/or if the quality of your work does not improve after our feedback, you will be
removed from our active list of transcribers and will be put on "pause".

This means that you will no longer receive job invitations.

That is why it is important to always read the feedback that our quality controllers give you.
send! Mistakes can happen, but we expect an effort from you to
you improve.

5. Tips and tricks

Organize your time

Especially at the beginning, allow more time than you think you will need.
to complete your work. It is important to submit all work on time,
We are counting on you to do it. We offer you some flexibility, and
We hope to be able to count on you in return.

Do you need to do research?

A quick search (on your chosen support) will help you if you encounter a
difficult word or one you do not know and that you hesitate about
the spelling of a term.

The proofreading
 17

Before submitting your transcript, read it carefully. Does it read like

fluid manner? (for the Clean read). Does this reflect the audio, in the
minor details? (for verbatim).

Ask the community

Beyond the FAQ on our website, you can join the community of
transcribers!

Everyone always starts from the beginning and you will always find someone.
to help you in our community.

You can contact them either in your language's Slack channel or in the channel
general.

Spell checker

Our Editor has a built-in spell checker that you can activate at the top.
right of your screen, or by using the shortcut Control + G.
 18

6. Problem solving

6.1 First, look for the answer yourself before on ourdrive

shared.

6.2 The main problems encountered with the jobs of

transcription.

Audio, what to do in case of...

Bad Listen carefully to the difficult passages at least three times, to

quality reduced speed (0.7) and at a high volume. If you do not
or still don't understand these passages, (because of a noise in
speech background for example), please notify us at the address
inaudible nextoperations@[Link] soon as a small
e this part of the speech is inaudible, please specify this
segment with :

[inaudible, start-end]

For example: 'If I block the microphone, no one will know'

what… [inaudible, 00:03:40-00:04:00].

The Job that If you have been assigned a file in the wrong language,
you have thank you for always contactingoperations@[Link].
summer Do not try to transcribe the file yourself, it will
assigned would take too much time, and the problem can be easily
is not avoid by saving the file again in the correct
not language.
in the
good
language

The voices of In the case where the voices of the speakers overlap, it is
speakers important to try as much as possible to transcribe the
himself different voices as soon as possible. However, when
superpo it is becoming impossible, you can specify it on the
sent segment with:

[Crosstalk, beginning - end]

For example: 'Stop interrupting me all the time...

[Crosstalk, 00:03:40 - 00:04:00]
 19

NOTE:When you contact the Perfect Planner or the operations service, please
always indicate the name of the job (JobID).
 20

Editor, what to do in case of...

Problems The Amberscript editor will automatically try to
of save each file regularly. However, it is
savage always good to click the save button from time to time
bad or other. If a backup issue arises, the editor will
the display an alert message.
transcri
If this happens: do not close the editor. Please keep your
ption
open tabs and browser and wait a little before
est
power to save the file again. If you close your
lost
tab during the save error, you will lose your
progression. If you wait too long and the file does not
is still not saving, please contact
operations@[Link].

In the event that a complete transcription disappears, thank you in

first place to try using the undo button to
retrieve your work. If that doesn't work, contact
operations@[Link], in the
In most cases, we are unable to recover.
your work.

The text Sometimes, the automatically generated transcripts are

confused jumbled. This means that the words are in the wrong
order. This is different from translations of incorrect words,
what happens in most transcripts.

For example: "the transcriptions are generated.

automatically with words generated by words
in the wrong order. Sometimes it gets muddled. Most of the
transcriptions are wrong, which means it will be this.

If this happens, please contactoperations@[Link].

They will ensure that new transcriptions are generated.

Internet/ Connection problems can be very frustrating and the

Navigate most of the time we waste a lot of time. Here is
eur some advice:
problem
1. Check if other websites are not working either.
months
If they are working, the problem certainly comes from your ...
internet connection.
2. Check if using another browser site can resolve your issue.
problem. Our tools are used to function at their best
on Google Chrome.
3. Delete your cookies or try with a private window
your browser.
4. Refresh the page (be careful, do not do this if the file has
backup issues).
 21

5. Contactoperations@[Link] you cannot

not to work at all.

NOTE:When you contact the Perfect Planner or the operations service, thank you
always indicate the name of the job (JobID).