Dictation and Transcription Tips

About recording quality

Transcribing essential, but poor-quality dictation

In some cases you may wish to transcribe poor-quality dictation because the content is essential.  Type-thing Services will review each case individually to let you know what we can do to provide the best quality transcription possible.  Poor-quality dictation includes those which is noisy, muffled, simultaneous overlapping conversations, and two or more speakers recorded at greatly different volumes.  In some cases we are able to digitize and enhance the audio to remove noise or clarify the speakers.  See information on this page about digital audio files .

In such cases we will work to understand how many "indiscernible" sections are permissible.  This work is billed at an hourly rate depending on the services needed.  Rush service for poor-quality tapes, if available, is billed at rates higher than normal rush because of the large effort and likely transcriber fatigue involved.

Options for transcription

About cost to transcribe

Determining your likely cost for transcription can be confusing.  You may be faced with costs based upon amount of material ($/word, $/page, $/line, etc.), time to transcribe ($/transcriptionist-hour, $/minute-audio), piece rate ($/tape, $/CD), or bid by the job as requested by the customer.   Different vendors may not provide the same costing method. Technical content, amount of editing, and quality of recording are also factors for a transcriptionist.  Type-thing Services will consider all of these methods and help you understand what will work best for you; however, we may only bid a job in a particular method depending on the material to be transcribed and the consistency of the material on the tapes or audio files.

See our rates page for Type-thing Services' specific rate structure.

Here is some information about cost to transcribe that you may wish to consider.

Audio length & time to transcribe

How long does it take to transcribe a tape or file? Typically it can take from two to six times the length of the audio to transcribe. This large range depends on the type of material, how fast people talk, clarity of the audio, number of speakers, clarity of the speakers. Most of the work that Type-thing Services has performed has taken 1.5 to three times the length of the audio.  Single-speaker or interview transcription with clear audio takes the least amount of time.  

This information may be useful if you choose to pay your transcriptionist per hour of labor.  If you choose to do this, realize that a seasoned transcriptionist that is a fast typist will produce more per hour of labor.

Audio length & transcript pages

Technical content

The cost to transcribe is also dependent on the amount of technical knowledge or editing required for transcription. Medical transcription typically costs a bit more because of the additional skill, tools, and references needed to ensure an accurate and usable transcript or medical note. Why is this?  Because a knowledgeable transcriber will produce a higher-quality transcript that requires less of your valuable time to edit or correct.

In a business transcript, the cost will be higher if the customer requires extensive grammatical corrections.  Costs may also be related to the amount of time required to service your staff with inquiries, special requests, and "stat" or rapid turn-around requests.

While legal transcripts of interviews are often transcribed like business transcription ($/min), court proceedings tend to have a wide variability in transcribed content.  Therefore we prefer to charge in $/page for this because it better reflects the actual work and amount of transcription involved.

Verbatim transcript & edited transcript

Often times this is a hidden cost or savings!  Many types of documents (reports, letters) are edited as Type-thing Services transcribes, so you're not receiving a verbatim transcript. You're obtaining a finished product or a document that requires less editing. This saves you money because you are not charged for the words, lines, or pages that are edited out of your transcript.  You are also receiving editing in the cost of the transcript.  Of course, you may want verbatim transcripts for some applications (interviews, legal proceedings, classes, podcasts, etc.), and Type-thing Services can provide these.

Note that some transcription companies only provide you a verbatim transcript.  In this case you're paying a lot more for what you're getting.  Why?  You're paying for content in your transcript that will be eventually edited out.  You also have to pay an editor or edit yourself, which is an additional cost.  Since Type-thing Services transcriptionists have secretarial skills, they save you these costs.

See Options for Transcription section above for more about this topic.

Cost per line or word

Cost per minute

Conversion summary

• 0.5 to 0.75 pages / audio minute for slow speakers, some single-speaker presentations
• 1 page / audio minute for  single speaker regular dictation
• 2 to 4 pages / audio minute for multiple speakers or fast-paced dialogue

• Usually 1.5 work hours per audio hour to 3 work hours per audio hour
• Up to 6 work hours per audio hour for difficult audio (that requires listening multiple times)
  

Rush or expedited transcription

Finding the best cost to transcribe

Why not use computer dictation, speech-to-text programs?

Sometimes we are asked why a person considering transcription should not simply use one of the new and improving programs/systems for computers that type while you talk or convert dictation to transcripts. These programs recognize your speech as you talk into a microphone and type what you say into a document. The promise is that one would save much money in transcription costs. In fact, a number companies have sprung up and have marketed specific systems for the medical and legal communities, particularly around Electronic  Medical Records (EMR). In addition, you will find that some EMR providers put digital dictation through a speech recognition system before it is given to a transcriptionists ring in a role called a "speech recognition editor."

The quick answer:

Most individual professionals should not yet use speech-to-text.  Not only are these programs not accurate enough, you must spend more time on three things:  (1) You usually must talk slower and spend time training the system. (2) You have to correct the errors the program makes.  (3) You have to be your own secretary - correcting the errors you make, know proper grammar, spelling, formatting, etc.  Most times spoken English is not as it will be written.  Transcriptionists take care of this for you.

Reality creeps in where large organizations are requiring staff, due to hope of lowering costs, to utilize speech recognition systems.  Often this is tied to a move towards electronic medical records where the EMR company bundles speech recognition as an "extra added value."  Type-thing can provide Speech Recognition Editing to assist in this process.

The longer answer is that choices should be a matter of cost and convenience.  If the total cost to the dictator or organization is less using such text-to-speech systems, then they should use them. Our experience is that these systems are not yet sophisticated enough to pay for themselves, and may actually cost professionals and hospitals more due to their ongoing time investment.  There is no doubt that in the future these systems will be excellent, but for practical dictation, these systems have a long way to go.  In a number of cases the hasty move to an immature technology lines the technology company pockets at the expense of the medical community.  We have seen and learned of examples where professionals take more time with this technology and see less patients as well as have lower quality notes.  Here's what we think:

Type-thing Services does not use speech recognition for its transcription work.  Why?  Because even if the accuracy of the process was fairly high (and it never is - we explore it periodically), we would have to listen to the whole audio to verify the transcript was correct.  On top of that, we would have to edit your spoken word to something fitting for your needs.  We often do that on the fly while listening to dictation.  All of this editing a recognition transcript can take longer than just doing it from scratch.

You might also be interested in a well-referenced article "Rest in Peas: The Unrecognized Death of Speech Recognition," by Robert Fortner, that argues there are multiple barriers for good-accuracy speech-to-text systems.  The show that accuracy has not improved in speech recognition software since about the year 1999.  For limited uses like dialing a cell phone it might work, but for transcripts there are problems. 

Studies have shown that speech recognition plus human editors are less efficient than a traditional transcriptionist. (e.g., "Speech Recognition as a Transcription Aid: A Randomized Comparison With Standard Transcription," J Am Med Inform Assoc. 2003 Jan-Feb; 10(1): 85–93.)  That document concludes "speech recognition did not improve the productivity of secretaries or transcriptionists."  In fact, this study said there was a loss in that speech recognition reduced efficiencies of medical transcription to 87 percent of a transcriptionist alone!   On the other hand, technology is likely to improve.

When making that decision consider the following points:

Costs more overall

Why should a highly-paid professional spend time sitting in front of a computer editing their text, continually retraining the program for new words and names? Time is money and, at least today, text-to-speech programs seem to take time away from the professional. Some systems allow you to talk into a dictation machine; however, you must still worry about the points below.

Talk clearly

You must talk clearly and enunciate each word. The programs are getting better, but you cannot slur your speech, talk extremely fast, etc. The surrounding must be quiet, not a noisy room, lobby, or car. Multiple people cannot be talking around you or in the dictation.

Do you speak like you write?  Quality transcription is more than speech recognition!

Transcription involves punctuation, grammar, and formatting at the least.  Spoken English is vastly different from written English.  You may be surprised at how unstructured spoken English appears when typed.  When requested, we regularly correct our clients spoken English into professionally appearing written transcripts.  With speech-to-text programs, you will have to train yourself to speak in written English form.  Some EMR systems require the medical professional to call out punctuation!  This is not the only problem.  Many of our clients do not always speak in an ordered linear format from beginning to end.  Part way through a dictation, they will remember something that should be inserted elsewhere in the final product.  With a human transcriptionist, you only have to give direction and the content for this to be accommodated.  With a speech-to-text program, you end up spending more time.

Are you a secretary?

Once you have your transcript in the computer, do you know all the rules of grammar, spelling, formatting, etc.?  If you do, great. Now waste your valuable time performing such an administrative function. Is your staff going to edit the transcript? If so, great. Do they have good secretarial skills that will produce letters, reports, and documents that present the professionalism you need?

Train, train, train...

Although these programs are getting better all the time, they are not yet like the science fiction portrayals of computers recognizing you talk. The programs today cannot recognize speech from every person; they must be "trained." Even after they are "trained," they will make occasional errors, and will almost always not understand uncommon words, new words, or new names. You must train them at least once each time such words arise.

Not for groups or poor quality recordings

Even if the computer can understand one person speaking clearly, it cannot yet even attempt to untangle a transcript of multiple speakers, sometimes talking at once, often in noisy conditions, some talking quietly, some talking loudly.

Transcription and Electronic Medical Records (EMR)

Is 24-hour Turnaround Possible or Reliable?

Is off-shore transcription worth it?

It's not a secret that, as with many industries, off-shore competition has moved in to challenge providers of transcription services in the United States.  Is it worth it to you?  While this answer is something you'll have to decide for your situation, make sure you're looking at savings in the bottom line cost of your transcription solution.  The following information may be of use in considering this option.

First of all, do you even know when your transcription is being sent off-shore?  Many of these companies have purchased domestic companies, ".com" web sites, or established offices in the U.S., but still send the work abroad.  You may interact with a U.S. citizen and call a U.S. phone number, yet your dictation is sent outside of the United States.  Make sure to ask where your work will be done, and in general by whom.

Total cost for your transcription is likely related to these four items.  Some items may be more important to you than others depending on your business needs.

1. Raw cost to transcribe ($/line, $/page, etc.)
2. Risk - Privacy of information sent outside the United States
3. Spoken versus edited text
   
4. Quality of the transcript or product (ability to use the product)
   
5. Customer service, responsiveness and flexibility to your needs
6. Specific off-shore issues (turn-around, privacy, security, export-control regulations)

1. Raw cost to transcribe

Because raw cost to transcribe is the initial attractive feature of off-shore services, your initial raw cost should be less.  You should understand that raw cost is not your total cost.  Consider the total cost in your decision.  Total cost may be affected by the following three items.

2. Risk - Privacy of information sent outside the United States

The basic fact that your work is sent off-shore may be an issue you've not considered.  One positive issue is that if you require quick turn-around, those working on the opposite side of the globe can transcribe while you sleep, so your work may be ready the next morning, in less than 24 hours.  There are a number of potential negative issues

• When you send your audio and resulting transcripts outside of the United States (with or without your knowledge), you are sending it to locations not covered by United States law.   If your information is private or covered by a number of laws to which you are held accountable, can you be sure you've performed due diligence in protecting that information?  If that information is disclosed, can you obtain damages from a company in a remote country, one you may not even be able to identify?

• Is the process to send the work abroad such that it meets your security and privacy needs?  Company proprietary information or health information (HIPAA) could be compromised.  It is not just the transmission of your audio that should be secure, but there should also be assurance that the companies and individuals abroad can maintain privacy and security.  Their networks, computers, and facilities should be as secure as domestic providers.  A number of instances in the press have shown that security abroad is an issue.  Even if they have excellent computer and information security, the people working there are under foreign government influence and different rules.  If something does go wrong, how are you going to take action against an off-shore company?

• A critical problem to consider is export-control regulation.  This appears to apply mostly to technical data, not necessarily personal medical information.  Export Administration Regulations ("EAR") and International Traffic In Arms Regulations ("ITAR") control the export of commodities, software, technical data and other information to foreign countries.  If you send information abroad in audio files which is covered by these regulations without the proper export licenses, you can be fined and go to jail.  If non-U.S.-citizens within the U.S. access this information, it is also considered an export.  Check with your company or institution to see if your transcription contains export-controlled information.

3. Spoken versus edited text

In many off-shore transcription services, you are charged for every word that you speak because your transcript is a literal copy, often inaccurately, of what you say.  With Type-thing Services, you are not billed for your redundant words and comments to our transcribers.  We usually reduce the size of the transcription product you receive because we edit it as we transcribe!  In addition, you or your staff must now spend time editing the transcript from spoken to written English.  So, you pay less because there are less words and lines and you have a higher quality edited product.  This is double the value!

4. Quality of the transcript or product

The most common complaint we've heard from clients that have tried off-shore services is that the innate language barriers cause inaccurate transcripts, grammar is poor, and there are spelling  problems.  If the pool of transcriptionists is large and transitory, your quality may be variable.  This is worsened by U.S. clients that tend to talk fast, mumble, or of have a strong local U.S. accent.  If you don't mind a poor-quality product, this may not influence your decision.  Just remember that a poor quality product may influence your total costs now because you have to fix the product yourself, or it may influence your future costs should you call upon the transcription in the future and find it useless.  If the transcript is a form of insurance or mandated record, you may be found negligent for accepting a poor quality transcript.  If a faulty transcript is used in the future, it may cause erroneous actions that will increase your costs.  Note that you can get poor quality from domestic sources too, so this is not just an off-shore issue.  Off-shore sources may be able to produce high-quality product if they have the right staff; however, they are having an increasingly difficult time finding qualified staff.

5. Customer service, responsiveness, flexibility

If the off-shore services and their domestic front offices cannot provide you with the customized and responsive services that make your work efficient, then that adds to your total costs.  If this doesn't matter to you because you need little customer service, then off-shore services may be more attractive.  Common complaints we've heard from clients include problems redressing quality issues, following up with updates, and corrections.  Because many off-shore services save money by having large-scale operations, they may also have some trouble at customizing their process to fit your business needs.

How does Type-thing Services know?

We receive clients who have not been satisfied by their experience with off-shore transcription services for many of the reasons noted above.  We have been contacted by numerous off-shore companies that have wanted Type-thing Services to front their services to U.S. customers. We have seen transcripts produced by off-shore transcription companies when clients were not happy with the results.  We have called to understand the utility of using such services ourselves.

Does Type-thing Services use off-shore transcription services?

No.   All our work is performed in the U.S.A by U.S. Citizens.  Most all of our work is performed nearby our location so that we know and can interact personally with our transcriptionist.  Quality is an essential element of the product Type-thing Services provides.

Is off-shore labor plentiful?

Not necessarily.  Plentiful qualified labor is the entire premise for off-shore transcription companies ability  to maintain low rates and quality.  Recent news articles show that as the global economy evolves, off-shore markets are experiencing difficultly in obtaining enough qualified labor for many technical tasks and service tasks that require training.  Their qualified staff must be paid more or they move to higher-paying jobs.  To maintain lower rates, they must use less-qualified labor.  The grass is not always greener on the other side of the ocean.

About digital audio files

With establishment of multimedia computers (audio, video, etc.) as the norm, more material is being generated in the form of digital computer files. Digital hand held dictation devices are now available that record to a memory card and can generate audio files you can place on disk or send over the Internet. Type-thing Services has the ability to convert and transcribe such files that come in a variety of formats.

We can also generate these files for use on your web site from your audio or video tape.  We'll work with you to understand what you need for your application.  Part of our service is understanding these formats and knowing which work well on the web and Internet.  We use multiple methods to make the smallest possible audio file for your purpose so that the file can be downloaded or transmitted most efficiently.  See our Web and Internet Services page for more detail.

• Tracks: Mono, Stereo, Multi-track
• The more tracks you have, the more file size is required.  Stereo or multi-track is not typically useful for transcription unless each track represents a separate microphone in a different location.  In that case, all the tracks can be combined for transcription or transcribed separately.  Courtroom recordings often have four separate tracks (judge, witness box, defense/defendant, prosecutor/plaintiff).
  
• Sample rate
• Sample rates tell you how many times each second the audio is recorded.  Faster rates have better quality but take more file size.  Slower rates have less quality but produce smaller files.
• Typical sample rates are resumed in samples per second and are typically 6000, 8000, 11025, 22050, 32000, 44100, 48000, 64000, 88200, 9600, and 176400.   CD-quality audio is 44100 samples per second.
• The frequency of audio you can reproduce in a digital file is at most half the sample rate.  So, at 44100 samples per second, a CD audio can reproduce at most 22 kilohertz frequencies.
• We recommend that for voice transcription you have a sample rate at least 22050 samples per second.  We can transcribe lower sample rates, but the audio quality decreases with lower sample rates.
  
• Compression
• Some formats of audio permit various degrees of compression, which makes the file smaller at the expense of audio quality.  Most of the time audio quality is not impaired, but at extreme compression it may be affected.  These file formats are known as "lossy" in that they can loose audio quality.  An example format like this is MP3.
• Compression is a trade off of file size to audio quality.  For dictation, select one that does not significantly impair audio quality.  MP3 files can compress more with the Variable Bit Rate (VBR) format, but don't use that because transcriptionists cannot use that directly.  Instead, use the Constant Bit Rate (CBR) format.
  
• Sample size (bits)
  
• Each sample taken typically has a fixed size, measured in bits.   The larger this size, the more accurately the audio can be reproduced and the larger the resulting file.  The smaller this size, the less accurate the audio, but smaller the resulting file size.
• Typical sample sizes are 8-bit, 16-bit, and 32-bit.  The most popular and size we suggest for transcription is 16-bit.
  

Which digital audio files should I use?

Which files are the best to use? 

Which type work on the Web and Internet?  The web and Internet use of audio is evolving.  For transcription, current influence is created by MP3 players, Apple's I-Pod, and digital dictation machines.  MP3 and WMA file types seem to be popular at this time.

Original sound files included the Next/Sun (AU extension) files and the also, due to Windows' popularity, the WAV files.   Later formats like Quicktime and Real Audio showed promise in reducing the file sizes and added ability to stream the audio.  Streaming means the audio is played over your computer's speakers pretty much it arrives.  Before that, the entire audio file had to be downloaded before it was played, which was inconvenient for large files or those that were transmitted in real time.  Now MPEG3 files are popular for music files and are very good at compressing audio as are WAV type TrueSpeech files.  The answer to the question really depends on what you are trying to do and what resources you have to provide the audio files to the user.  Some issues include:

• How are you going to provide audio files to the users?
• Will the users be able to work with the audio files you provide?
• What bandwidth Internet connection do the users have?
• Are the files going to be downloaded or streamed?

• Newer audio file technologies typically make smaller files.
• Some file formats (or options within a format) can reduce size.  This is compression.
  
• As the number of samples per second is decreased, so is the file size (usually).
• As the number of bits of resolution (dynamic range) per sample decrease, so does the file size (usually).

What can be done with Audio files to edit  the recording?  Digital audio files can be easily edited to produce a good quality finished product. For this discussion, editing is the simple rearrangement of audio segments that is analogous to cutting and splicing audio tape.  Some examples are:

• Audio can be easily deleted.
• Audio can be easily moved, copied, or spliced.
• Silence can be added or removed.
• Audio from other sources can be spliced into the recording.
• Multiple tracks can be converted into one track.
  

What can be done with Audio files to enhance the recording?  Digital audio files can be enhanced either to improve poor-quality sound or by adding various special effects.

• Uneven speaker volumes can be adjusted so low volume speakers can be heard.
• One speaker can be increased or decreased in volume to generate a sense of distance or depth.
• Many constant background noises (hum, buzz, noise, etc.) can be eliminated without distorting the speech.
• A large number of recording studio special effects can be added to all or parts of the recording.

About audio files from video

About "tapeless," digital," and "phone-in" dictation

These approaches to dictation and transcription have become the norm in the industry.  "Tapeless" and "Digital Tapeless" are becoming archaic terms for Digital refer to dictation without audio tape. This could be a hand-held recorder that stores your dictation in memory modules, or it could be a phone-in dictation system.  These types of devices have essentially replaced hand held tape recorders.

First-generation digital dictation units (popular types by Sony, Olympus, etc.) typically produce audio in proprietary formats that are difficult to convert without their own proprietary software.  Newer devices coming out after 2009 started to create files in standard file formats such as MP3, MP4, and WMA.

Type-thing Services prefers you consider phone-in dictation because of the numerous advantages it offers. See the "Phone-in Dictation" page on this Web site.

About audio tape sizes and formats

1. Micro cassette,
2. Regular cassette, and
3. Executive cassette.

Shown above are the regular cassette (top), executive (left), and micro (right) with approximate sizes for each tape. Micro and Executive cannot be used in the other's machines. Executive tape dictation systems are more expensive but provide superior clarity of dictation.

Formats

Most popular recorders use a single track of mono or stereo audio. Some of them have two speeds that you can record your audio. Recording on the fastest speed produces higher quality dictation, but provides less recording time on the tape.

Multiple-track recorders are typically used in settings that require very accurate transcriptions and have multiple persons that might speak simultaneously. For instance, courtroom transcripts are often taken by a four-track recorder with each person wearing a separate microphone and recording on a different track of the tape: judge, two lawyers, witness. Multiple-track recorders are rare outside of the courtroom setting. However, they provide superior transcripts because the transcriber allows one to listen to each track individually or all tracks at once.   Again, digital dictation systems have primarily replaced tape-based recording.

About quantity of dictation per tape

 email: michele@type-thing.com
web: http://www.type-thing.com/