Casting Words Style Guide PDF

Preview

Summary

Download Casting Words Style Guide PDF

Description

1 THE BASICS 1.1 CONFIDENTIALITY Confidentiality is of utmost importance. Do not ever post or share any transcripts or audio that you receive from CW, or create for us, and never leave these materials anywhere that they will be accessible to anyone besides CW and yourself.

Inaccuracy is the worst problem for a transcript to have. Inaccurate work will not receive a bonus, may be rejected, and we may reduce your badge. 2.1.1 WHAT TO DO WHEN YOU CAN'T UNDERSTAND A WORD OR PHRASE See the Tags section, below, for complete instructions on how and when to use them.

You may want to save the text files from your completed jobs for a short time after submitting, just in case of questions or technical issues. It is fine to hold onto them for a few days, but all text and audio files should be deleted in a timely fashion.

Your transcript won't be downgraded for using [xx] or [?] for words that are legitimately hard or impossible to understand. However, there are several things you can check or try before marking something as inaudible:

1.2 TRANSCRIPTION REQUIREMENTS

Check the job page The customer's special instructions, the list of names given, and the title of the assignment can all contain clues as to the names and spellings of individuals and organizations mentioned in the audio, as well as the fields of work or study they may discuss.

It must be accurate The words you type must be the words that are spoken. It must be written in standard US English Do not use netspeak or phonetics, such as "u" for "you." It must not paraphrase Do not rearrange words. Do not correct the speaker's grammar. Do not cut words that you think are irrelevant or off-topic. Do not include any words that are not spoken. It must not include anything that wasn't in the audio Do not include titles, job numbers, page numbers, or your own comments along with the job. It must be formatted correctly See the relevant sections below for full details on sentences, paragraphs, tags, and speaker labels. Transcribers never add timestamps. Non-verbatim jobs must be "cleaned up" Non-verbatim work must be lightly edited to remove stutters, filler, and false starts. Our standard transcription jobs are all non-verbatim. Verbatim jobs must be true verbatim Verbatim work must include every utterance and sound event, exactly as you hear it. Transcribe stutters as accurately as possible. Do not leave out any filler words unless specifically directed to do so in the "Notes" section of the job. See the Verbatim section, below, for full details. Our verbatim jobs are clearly marked as being verbatim.

2 MOVING BEYOND THE BASICS

Use clues in the audio Pay attention to information contained in the audio itself. If an inaudible word is used elsewhere in the audio and you get it, you can then go back and correct the earlier instances. Likewise, if the interviewer concludes by saying, "Thanks, Dave!" and the interviewee's response indicates that he is Dave, you can go back and label the interviewee as Dave throughout the transcript. Do some basic research If you can't understand a word but you know the context (for example, it's the last name of a company CFO, a drug used to treat asthma, a city in Ireland) you can often figure it out with a couple of searches by looking up relevant keywords, together with your best guess at the spelling of the unknown word. 2.1.2 DON'T MAKE UP WORDS The speakers in your transcripts can use their own made-up words, and you can transcribe them as spoken. If a speaker says "ginormous" or "smartification," that's no problem. However, transcribers and editors should not make up words. When we say "Don't make up words," we mean two things: Don't spell words phonetically. All words should be spell-checked and must be

2.1 ACCURACY AND COMMON SENSE

actual English words unless, again, the speaker is deliberately using made-up words. Your transcript

will be heavily downgraded for using phonetic spelling, such as "coco van" for "coq au vin."

2.2 WHAT TO CUT: REMOVING FILLER AND FALSE STARTS

Don't include words that make no sense in the context of the audio Read your transcript before submitting, as if you were reading an article or story. The words you transcribed may sound like the syllables that were spoken, but if they don't make sense, it's probably not what the speaker was saying.

Spoken language differs from written language in several ways. Many speakers pepper their speech with "um" and "uh," small stutters, false starts, and filler words, especially in informal conversation.

For example, a TSA officer is unlikely to mention "hot gowns" when discussing his work. It's more likely "pat-downs," since that's what he probably does all day. 2.1.3 RESEARCH AND VERIFY Transcribers are expected to do the legwork when it comes to verifying spelling, format, and punctuation of any proper names or unusual words. Leave a comment to confirm that you have done the research. If you can't verify something, be sure to tag it with [sp] or [?], and again, leave a comment so the editor knows you were unable to confirm. Verify spelling Look up proper names Some company, organization, and brand names are hyphenated, some are not. Many use unusual spelling, spacing, capitalization, or punctuation. Follow accepted conventions Got a phone number? Look up the standard notation for phone numbers in the country in question to find out whether and where to use spaces, parentheses, dashes, or other punctuation. Twitter handles start with an @, so if someone is talking Twitter it's @GeorgeTakei, not "at George Takei" -- and of course you've already done a quick search to make sure you're spelling "Takei" correctly. URLs and email addresses should be written in lowercase unless the speaker specifies caps. Write them out in their final, usable form, even if the speaker stops and spells something in the middle. If the speaker says, "Amazon, with a Z, dot com," just write amazon.com. If the speaker doesn't say "http://" or "www," don't add it in. In a few cases, the URL doubles as a company name, such as Amazon.com. If it is being used as the company name rather than as a URL, capitalize appropriately.

Unless the customer has ordered a Verbatim transcript (in which case see the separate section, below), these verbal excesses should be trimmed so that the final product reads smoothly. This section applies to non-verbatim jobs, which make up most of the work at CastingWords. 2.2.1 WHAT NOT TO CUT You will notice that all of these guidelines are about trimming "fluff" that does not add meaning to the transcript. Do not skip content just because the speakers have wandered off topic. Our job is to transcribe what is said in the audio, not decide which parts are important! 2.2.2 CLEAN UP FILLER AND NON-WORDS

Non-words Remove non-words such as "ah," "er," "um," "uh," and "mm-hmm," unless they are absolutely necessary to indicate meaning. If "mm-hmm" or a similar utterance is the only answer to a direct question, then it should be kept, or information will be lost.

Filler Any word or phrase that a speaker uses habitually and repetitively, which does not contribute meaning to the sentence, is likely filler. Some common ones include "like," "you know," "really," "kind of," "so," and "OK." Some speakers start every sentence with the same word or phrase, such as "OK, so..." or "Yeah..." That's filler, and should be cut. An interviewer may acknowledge an interviewee or prompt them to continue by murmuring or saying "Right," "Yeah," or "Mm-hmm" whenever they pause. That's filler, and should be cut. Some speakers, especially in business settings, have a personal catch phrase such as "essentially" or "quite frankly" that they sprinkle into nearly every sentence. That's filler, and should be cut. At the same time, we want to preserve the "flavor" of a person's speech, so it's OK to leave in an instance or two of a catch phrase, especially in

places where it seems to actually mean something as opposed to being solely a verbal tic. 2.2.3 CLEAN UP STUTTERS AND FALSE STARTS Remove false starts unless they add information that is not included elsewhere. Also remove stutters and words that a speaker repeats while thinking or searching for the right word. Use common sense! Sometimes, the participants in an audio are recording multiple takes of an interview or presentation that will eventually be edited into a seamless, polished final product. You can clean up minor "conversational" false starts in that case, but obviously not every failed take is a giant false start to be deleted! If the speaker trails off or flubs a line and remarks on it, all of that should be included. Start a new paragraph when the speaker starts a new take. Examples If the speaker says: What did I do with the dog's...I need to get to the bank before it closes You should write: What did I do with the dog's...I need to get to the bank before it closes The false start is kept because it includes information that is not found elsewhere. If the speaker says: I really need to...need, um...I've got to get to the bank before it closes You should write: I've got to get to the bank before it closes. The false start is cut because it adds no new information.

3 TRANSCRIPT FORMATTING In most cases, the TC (Transcript Chunk) you work on is just one part of a larger audio file, which in turn may be part of an order containing multiple files, and a customer may place many such orders over a period of time. It's important that we deliver consistent results, from one file to the next, and one order to the next. 3.1 SENTENCE AND PARAGRAPH STRUCTURE 3.1.1 THE BASICS Tip: You can adjust the settings on your word processor of choice so you don't have to tweak your text every time you submit.

Use Word Wrap If you work offline (which is usually a good idea), just make sure there are no stray line breaks in the middle of your paragraphs after you paste your work into the job page. If there are, fix them before submitting. Don't double space after sentences Start every sentence with a capital letter and end it with punctuation This includes the first word of a quoted sentence. It also includes the first word after an ellipsis, if -- and only if -- it is the beginning of a complete sentence. Also capitalize and punctuate words and phrases that stand as a complete statement (such as "OK") even if they are not actually complete sentences. Break compound sentences into smaller ones where possible Shorter sentences are preferred, as long as the resulting sentences are still grammatically complete, and not sentence fragments. If a sentence is ridiculously long (more than 400 characters), go ahead and break it into fragments. Keep paragraphs short Keep paragraphs to a maximum of 400 characters. Put a blank line between paragraphs Separate paragraphs with a blank line. Start a new paragraph at every speaker change, and at logical points if a single speaker goes on for more than one paragraph. Tags that stand on their own, such as [laughter] or [crosstalk] are treated like paragraphs as well, with a blank line above & below each one. Do not indent anything Do not insert spaces or tabs at the beginning of any paragraph, before speaker labels, or before a tag that stands on its own line. 3.1.2 CONJUNCTIONS Conjunctions such as "and," "because," "but," "or," "so," and others are generally used to join two parts of a sentence together. They therefore also provide a natural breaking point that can allow you to create two shorter sentences rather than one long one. Whenever possible, avoid starting sentences with conjunctions, especially "and" and "so." You can often simply cut an "and" or "so" from the beginning of a sentence, without changing the meaning. If cutting the conjunction out would alter the meaning of what was said (be especially careful

with "but" and "because"), you may be able to leave the longer compound sentence, rather than breaking it into two shorter ones. Conjunctions should only be left at the beginning of a sentence if neither of those options is workable for a specific case. 3.2 SPEAKER LABELS Speaker labels should be useful to the customer Speaker labels should be formatted correctly Note: We use the phrase "speaker labels" rather than "speaker tags." Only items in square brackets are considered "tags" in the CW system. Tags and labels have different formats, and serve different purposes. 3.2.1 THE BASICS What is a speaker label? A complete speaker label includes the speaker's name (or role or other identifying label), followed by a colon and a space. Each word in a speaker label should be capitalized. Example: Interviewer: Hello, thanks for coming today. Woman 1: It's a pleasure to be here. A note on formatting In some views, the system will display the speaker labels from your transcript in boldface. Transcribers and editors work only in plain text, and are never required to apply bold, italics, or other formatting to submitted text. Occasionally, a speaker label may contain punctuation or special characters such as accented letters. The system may not bold these labels as expected in some views, but such labels are still valid, and the copy of the transcript delivered to the customer will display them correctly.

If you don't know speakers' names, or can't be sure which name goes with which voice, use roles instead. Examples include Interviewer, Interviewee, Lecturer, Student, and Moderator. Use gender (Man 1 and Woman 1) only as a last resort, if you don't know anything else about the speaker. Using names in speaker labels A speaker's full name, sometimes including a title, may be found in the audio, or may be given in customer comments or on the list of names included in the assignment. The first time the person appears in your audio chunk, use their full name (including title, if any) if you have it. NOTE: After the first mention, you will shorten it to the person's first name. However, in a formal setting, it may be more appropriate to use last names with titles or honorifics instead. Example: Dr. Jane Michaels may be shortened to Dr. Michaels, if the event is a panel discussion where speakers are formally introduced. Pastor Linda Thomas may be shortened to Pastor Linda, if she, her congregation, or her website refer to her that way. Using roles in speaker labels When names are not known, you should make speaker labels as informative as possible by identifying speakers based on their roles in the audio. Some examples include: Interviewer, Interviewee, Lecturer, Host, Student, Facilitator, Congregant, Audience Member, Passerby, Announcer, Tech (for camera or sound crew).

How often do I label the speakers? Label the first person who speaks, even if they are the only speaker in the entire audio.

Some speaker labels are too general or too informal. Unless specifically requested in the customer notes, you should never use speaker labels such as Respondent, Speaker (or Speaker 1, 2, etc.), Female, Male, Lady, Dude, Person, or People.

Thereafter, insert a speaker label at every speaker change, and after any tag that stands on its own line (such as [laughter] or [crosstalk]), even if it is the same person who was speaking before.

When to mention gender In the absence of a name or role, if all you know about a speaker is their gender, use numbered "Man" and "Woman" labels.

If one person's speech is broken up into multiple paragraphs, do not re-label the speaker at the beginning of each paragraph.

If there are children in the audio and you can't tell if they are boys or girls, you can use numbered "Child" labels instead.

3.2.2 CHOOSING THE BEST SPEAKER LABEL

Unless specifically requested, "Male" and "Female" are never valid speaker labels on their own (use "Man" and "Woman" instead). However, you should apply "Male" and "Female" as modifiers if (and only if) people of different genders have the same role in the audio. When to number speaker labels Transcribers should always use numbers with "Man" or "Woman" labels, even if there is only one person with a "Man" or "Woman" label in a given transcript chunk. Editors should remove the numbers if there is only one person with a "Man" or "Woman" speaker label in the entire audio file. With all other speaker labels, use numbers only if two or more people have the same gender and the same role. Examples A podcast has two unnamed hosts of the same gender. Use numbers only: Host 1, Host 2 An unnamed man and woman conduct a joint interview. Use gender only: Male Interviewer, Female Interviewer Three unnamed students participate in a study. Use both gender and numbers, as needed: Female Student, Male Student 1, Male Student 2 3.2.3 SPEAKER LABELS AND AUDIENCE PARTICIPATION When a whole audience speaks together, they are labeled "Audience" unless they are gathered in a church or other place of worship, in which case they are a "Congregation." A single member of the group is an "Audience Member" or "Congregant." Don't worry about telling audience members apart. Each one will just be "Audience Member" or "Congregant," with no numbers or gender needed. Please note that not every large group is an audience. "Class," "Group," or other labels may be more appropriate, depending on the context of the audio. Use common sense! 3.2.4 SPEAKER LABELS FOR GROUP DISCUSSIONS When to use Group Style In groups with many active speakers, it may not be possible to track each one.

The following rules apply to any job that is identified as Group Style on the job page, or that is described as a "Focus Group," "Panel," "Round Table," or "Group Discussion" in the tags, customer comments, or job title. If you think a job should be Group Style, but it is not identified as such, check with Support if possible. If you can't get an explicit go-ahead, use common sense and make the decision yourself. If it is legitimately a group discussion, focus group, or other type of meeting with many participants, Group Style is probably justified. Files with four or more active speakers may qualify, especially if they are all the same gender. If you can track the individual speakers, then do so. If you have a main speaker (or several) with audience participation, see the section on audiences, above. Audience participation is not a case for group style. If it is a seminar setting where there are many speakers but they are each introduced and speak individually, that is not a case for group style. Use your best judgment and always leave a comment if you feel a job must be done in Group Style, but you can't get the official go-ahead. Using group format speaker labels on a job that does not require them is cause for rejection. Speaker labels on Group Style jobs The only person to be tracked in a group discussion is the group leader, if any. Use the facilitator's name if you know it, otherwise label that person "Facilitator." If there are two identifiable facilitators, use their names if known. If you don't know their names, use "Facilitator 1" and "Facilitator 2" if they are both men or both women. Use "Male Facilitator" and "Female Facilitator" if there is one man and one woman. All other participants in a group discussion are labeled Female Participant or Male Participant, with no numbering. That's the whole point of group format -- not having to track individual speakers. If the participants go around the group and introduce themselves, then use their names for that part only, then switch to the generic, nonnumbered labels thereafter, even if you can sometimes tell who is saying what. 3.3 TAGS

3.3.1 THE BASICS Tag Format CastingWords tags are only ever formed with square brackets [], never parentheses, curly brackets, angle brackets, or other symbols. Text in tags is lower case except in the (rare) case of a proper noun being included in a tag such as [nonEnglish speech]. Tag Placement A tag relating to an individual speaker's words or actions is placed inline with the rest of the speech, as if it were a word. Examples include [xx], [?], [sp], and [laughs]. Use a space between these tags and the words ar...