Human vs. Machine: Should You Use a Manual or Automated Transcription Service?

Speech and voice recognition software is getting better than ever.

In the age of Siri, Alexa, and Google Assistant, we expect more advanced transcription software. However, are any of the automatic transcription services good enough for your audio or video?

How many times have your automatic transcriptions included potentially grave (and often tragically embarrassing) errors?

Transcribing words is relatively easy for machines, and they’ve gotten really good at it. Just take a look at this chart by Kleiner Perkins.

What’s difficult is to infer what you truly mean by saying these words, which makes it more prone to error. Humans, on the other hand, are much better at discerning and recognizing the context of a conference call, webinar, podcast, or any kind of content.

Transcripts produced automatically are quick and private. Yet, they eventually need intervention by hand for proofreading. Others who are looking for accuracy and convenience think that they might as well order manual transcription services.

If you’re unsure whether automatic or manual transcription is the better solution, we’ll be comparing and discussing the pros and cons of both options in detail.

Find out whether you should purchase a manual or automatic transcription service…Human or machine?

The Benefits of Automated Transcripts

First, let’s go over the benefits and downsides of each option, beginning with automatic transcription services.

Transcription tools like Scribie, AWS Transcribe, or The Cloud Speech API by Google converts speech into text transcription from audio. Sometimes, you don’t even need a recorded file. Scribie, for instance, lets you directly make in-app calls where conversations are recorded for automatic transcription.

Automatically turning your video or voice file into a transcript is fast and easy. The average time it takes Scribie to transcribe a 2-hour file is 30 minutes. You’ll have a fast, first draft of your recording in no time.

Besides speed, it also comes at a lower cost for those whose needs don’t require a perfectly documented version of their audio.

Concerned about privacy and confidentiality? With automatic transcription, no other person would listen to your audio. Although some platforms, like Google and AWS, transfer your data to Google servers or others. Scribie is unique in that we have our own tech for privacy concerns and don’t require 3rd-party platforms to process your recordings.

Is It Good for You?

That depends on how you’ll use the transcript. If you’re simply using it to refer to while listening to the recording, you’ll have an easier time processing the information and content of the recording.

Transcripts of interviews, video content, and raw footage are often used by media outlets and journalists as a first draft. Scribie also provides automated subtitles along with the transcript. Users will be able to download the file, as well as upload it to their YouTube account and add it as a caption.

Researchers whose files may contain highly-technical terms or crucial data might prefer automatic transcriptions. Later, they’ll be able to edit the copy themselves for preciseness and accuracy.

For highly sensitive transcripts, such as psychotherapy sessions, rest assured that no other person would listen to the audio or watch the video.

Many businesses also use automated transcripts as a backup for detail discussed in a meeting. It’s also useful for contact centers who record a high-volume of calls with customers.

Automated transcripts are Ideal for:

  • Journalists
  • Medical
  • Research
  • Businesses
  • Contact Centers
  • Media

Get a fast and easy transcript of your recording. Upload a file on Scribie for automatic transcription now.

The Downsides to Automation

Transcription software doesn’t have language inference systems. Without this tech, the machine won’t be able to make corrections based on context, as well as identify different speakers, appropriate paragraph breaks, punctuation marks, and more.

Scribie, however, has a highly-advanced speech recognition engine tuned specifically for transcription of all file types. Unlike others tuned only for conference calls, Scribie can automatically recognize multiple speakers, whether on the other side of the call or sitting right next to each other.

If you find transcription technology that guarantees 100% accuracy, they’re either overshooting claims or over-optimizing for one particular scenario.

Of course, accuracy will also depend on certain factors such as the speaker’s accent or whether the audio is of high quality.

Typical Errors in Automated Transcripts

There are a few usual errors you can expect based on the limitations of today’s transcription automation software.

Artificial intelligence has difficulty detecting speaker turns, especially when there is fast back-and-forth between multiple speakers. All systems have a 20% to 30% failure rate for fast turns. In the most ideal scenario, a system will be able to figure out 90% of the turns. This means paragraphs won’t be broken correctly, words will be missing, and speakers won’t be tracked.

Names of people and places are also often mixed up. There is a village in Wales, for example, called Llanfairpwllgwyngyllgogerychwyrndrobwllllantysiliogogogoch, which will most likely never be recognized by transcription software.

Source: Imgur

You can also rule out an automatic transcript from accurately displaying single or double quotes. A voice recording with “He said,…” will show the start of a quote but can’t determine where to close them.

Punctuations, such as hyphens, quotations, and exclamation marks, as well as proper capitalizations, are often unrecognized by systems. AWS, Google Web Speech, and Speechmatics don’t support quotations.

Fortunately, Scribie has an 80% accuracy rate for punctuations. We support all types of punctuations, including quotes and hyphens. Our speech recognition engine also uses Continual Learning to consistently decrease the word error rate (WER). In five to ten years, we’ll be able to reach a 95% accuracy rate for most audio files with good quality.

Automatic transcription is usually the first step to transcription of data and the first draft is rarely the final one. As Scribie keeps your transcript synced to audio, you’ll have an option to manually edit the transcription afterward through our built-in editor tool.

The Benefits of Manual Transcription Services

Transcription by hand is great for a high-quality text version of your file, with an accuracy rate of anywhere beyond 95%.

Nothing is more convenient than handing over the task of transcription to a specialist. Manual transcription services can be beneficial for large files, fast talkers, speakers with accents, and audio challenges, like background noise or distortion.

Manual transcriptions can also be secure. When you use Scribie, your covered by an NDA. Our transcriber directly works through our editor and won’t be able to download the file.

A time-coding feature for manual transcripts helps you determine when a sentence is spoken, so the subtitle shows up at the right time.

Is It Good for You?

Hire a manual transcriber for a clean and polished copy that is presentable to audiences.

Brands and enterprises might need high-quality transcripts to remain compliant with brand guidelines. TV Show subtitles, for example, can benefit from a transcription by hand.

Some content, such as medical and scientific papers are high stakes. It might be preferable to have a human type out the findings. Manual transcription is also crucial for court transcripts, depositions, and other legal documents.

Despite the higher cost of manual transcription vs. automated, a human with attention to detail can save you money. Mistakes by automated transcription software can be costly. A missing punctuation mark once caused a rocket’s engine failure and ended up costing $620 million.

Manual transcripts are ideal for:

  • Enterprises and Brands
  • Research
  • Media
  • Education
  • Marketing
  • Legal

How Will You Use Your Transcript?

Deciding between a human or machine mostly boils down to the application of your transcript.

The modern tech of transcription software is impressive, with an accuracy rate higher than ever.

If you prefer a low-cost alternative and a more hands-on approach, you can use platforms like Scribie to quickly turn your recordings into written text which you can then edit afterward. Of course, doing-it-yourself may be impractical for large files.

Manual transcription services ensure your transcript is high-quality, convenient, and error-free for public or internal use.

Try Scribie’s manual transcription service by 100% humans. Order transcripts online today.

Easy Webex Transcription: Record & Document Your Teleconference with Scribie

Easy Webex Teleconference Transcription Service

Teams can be productive without having to be physically there.

However, are you documenting your important teleconferences?

Webex has been a key tool in the corporate setting. You use conference calls to stay in touch with remote colleagues, run workshops, or attend training events.

Teleconferences also allow you to connect with clients. You can easily hold sales meetings with potential overseas buyers or easily walk clients through solutions with consulting calls.

What happens after the conversation, though?

To get the most out of your events and meetings, you want to get transcripts out of your recordings.

Transcripts help you keep track of discussions and agreements made during the call. With a documented record of every word, you’ll have an accurate and precise reference point to look back on for future meetings. Plus, you can share the transcript with other team members who’ll be able to keep tabs on discussion points, gain insights behind decisions, and understand the next steps.

Sales and account management teams, for instance, can use a transcript for identifying members’ strengths and improvement areas when it comes to making clients happy. They can pinpoint what works and what doesn’t, and ultimately understand what it takes to close a deal.

Marketers and subject matter experts would also find teleconference transcripts incredibly useful. Workshop attendees would appreciate subtitles as they go through the recorded video afterward. For those who were unable to attend, marketers can offer the recorded transcript to go over the webinar’s content. Hosts can also use the transcript as speaker notes for future webinar sessions.

Alternative Ways to Generate Transcripts

If you’re exploring how to record Webex meetings, there are various ways to process and transcribe the increasing amount of video and audio data.

TIY or Transcribe-It-Yourself

First, you can do it manually yourself. It’s definitely tedious but not impossible to work through, especially for quick meetings. This, however, takes a lot of rewinding, slowing down audio, and close listening. Huge plus if you have a rapid-fire typing speed and can easily decode accents.

In any case, you have to review how to best spend your time and what it costs for you to get up to speed with this task.

Delegate to Team Members

Another option is to delegate the task of transcription to team members. This, however, can distract them from their current focus and more immediate tasks. Based on their strengths, their transcriptions might not be as accurate as you want.

Use Automated Transcription Tools

If you have zero bandwidth to manually transcribe, automated transcription tools are at your fingertips.

Important note: To use transcription services, you will have to convert your ARF file into a widely supported format like MP4, WMA or WAV.

Apps like these are fast and easy with a fair accuracy rate. Text is computer generated, using speech-to-text technology that ideally produces transcripts faster than the length of your audio. It could require some editing after, but still enables you to work faster.

You can try Scribie’s automated audio/video transcription here.

Just keep in mind that you will have to review any automatically generated transcript’s accuracy. For the best combination of cost, convenience, and accuracy, you can order transcripts that are manually written by our professionals at Scribie.

How to Skip the Webex ARF File

When it comes to transcribing Webex calls, you typically have to deal with ARF files.

If you’ve ever attempted to turn your recording to text, you might have stumbled upon Webex’s proprietary file format. Being exclusive to Webex, you’d find difficulties in terms of its closed nature and lack of portability.

It also involves hassles, like installing software to specifically handle this file format.

Converting an ARF file to a compatible format sounds foreign and a bit daunting. You’d rather just stick to familiar cross-platform file types such as MP4.

For clarity: ARF stands for advanced recording file. It is the default file extension format for Webex files. Within a .ARF, you have your recorded online meeting. This includes an index of the attendees and other details.

Other Webex files may come as a WRF. This is a local recording that’s made whenever the user directly records the meeting using Webex’s Record on My Computer option. ARF, on the other hand, are network-based recordings created by the meeting host using the Record on Server option

While ARF may come with advantages, its complications may hold you back from documenting your important teleconferences.

Fortunately, Scribie covers Webex transcription without having to work with ARF files.

By using our phone call service (no need to pay for the call), we’ll be able to record the phone calls and seamlessly deliver transcripts.

What’s great is that it also works for any teleconference with a dial-in number, including UberConference and Google Hangouts Meet.

The Best Way to Transcribe Your Webex Calls: Use Scribie’s Call Recording Feature

Scribie can handle your Webex recording and transcription in a single step. You can dial in directly on Scribie and we’ll record the video and turn it into text as well. Recording is free of charge. All you have to pay for is the call, with a flat rate of 3¢ per minute.

The recording is done on the server side and the audio data is captured directly off-the-wire. The recorded file sounds exactly as the actual call without any artifacts.

Here’s the very simple process:

Step 1: Go to Scribie’s One-on-one Call Service

Step 2: Fill in the required fields

Under Participant 1, add your number.

Add your Phone NumberUnder Participant 2, add your Webex dial-in number. If you’re using other teleconferencing solutions such as Google Hangouts Meet or UberConference, you can use their dial-in numbers too!

Add the Teleconference Dial-in Number

Step 3: Click “Start Call” to get Scribie’s service to call your phone number and the dial-in conference number.

From there, take the phone call as you normally would.Take the phone call as you normally wouldStep 4: Voila! Get the transcript in your inbox.

This easy method allows you to record and transcribe teleconferences without dealing with proprietary file types and extra software.

Over To You

Transcribing and recording Webex sessions can be cumbersome and time-consuming. Usually, you’ll have to record the calls, then convert them into a usable file format for transcription, and finally, upload them to be transcribed.

With Scribie, you can dial in directly and seamlessly get the transcript–all while still using Webex or any other teleconference service with a dial-in number. Transcription just got a whole lot easier.

Got a teleconference coming up? Take your call using Scribie with built-in recording and transcription.

New Feature: Automated Subtitles

We now provide subtitles along with the automated transcripts for $0.25/min of the audio. We support the SRT and VTT formats. You can order the subtitles from the button as shown in the screenshot below.

That will lead to the invoice page where you can make the payment. After the payment, the subtitle file will be replaced with the links to download the files.

The SRT and VTT buttons will download the file. The YouTube button will upload the file to your YouTube account and add it as a caption.

Try out our free automated transcription service with subtitles today!

New Feature: Priority Automated Transcripts

We are happy to announce that you can now prioritize the free automated transcripts and get it ASAP for as low as 10¢/min. Automated transcripts require a lot of CPU and GPU power. Therefore we queue these files up and process them one by one. Our processing queue is a FIFO queue, First In First Out. So new files are added to the back of the queue.

Sometimes our queue gets backed up and it may take a long time one particular file to reach the front of the queue. With this feature, you can pay to get in front of the queue. Your file will be pushed to the front and the processing will be started as soon as the current one finishes. Click the Redeem Now link and make the payment to prioritize. Here’s a screenshot of how it looks in your account.

 Try out our free automated transcripts today!

How To Correct Automated Transcripts With Our Editor

We provide a browser based-editor which can be used to quickly correct the automated transcripts. Click the Edit Transcript button to launch it.

Scribie Editor

The first thing you will notice is the audio waveform at the top. That is the audio player.  Clicking anywhere on it will take you to the corresponding word in the transcript.

The first row of buttons are the controls. Each button also has a corresponding keyboard shortcut so that you don’t have to use the mouse which saves a lot of time. The important shortcuts to remember are CTRL+P to play/pause and CTRL+O to rewind (CMD for Mac).

The second row of buttons are some controls for the text editor. Hover the mouse over the button to get a description of what the button does. It’s mostly self-explanatory.

You will also notice some text underlined in blue and red. The red ones are spelling mistakes. Run the spell check to correct those. The blue ones are where our speech recognition engine was not confident enough and so those may be mistakes. You can right click on those and choose Play Word to check the corresponding audio.

The following are the list of corrections which tend to be required in the automated transcripts:

  • Mistakes: These are words which are incorrectly transcribed. Most of these words will have blue underlines.
  • Speaker Turns: Our speech recognition engine misses around 40% of the turns. So some paragraphs may actually have two speakers in them (we are working to improve it).
  • Punctuations: There may be some missing periods. The commas and other punctuations are mostly correct, although we only provide the start quote. The end quote has to be manually inserted.
  • Capitalization: Some of the capitalized words may be wrong. Some other words may need to be capitalized.

We recommend the 2-pass approach to make the corrections. First play and check the blue underlines. Those are the low-hanging fruits and you can get them out of the way fast.

Next, play the audio from the beginning and make corrections as you go along. Whenever you notice a mistake, pause, make the correction, and resume play. Rinse and repeat till you reach the end of the file. Increasing the playback speed can also help in cases where the accuracy is more than 80%.

Once you are done with the edits, Click the Download button at the bottom for the Word Document or other formats.

Editor download files

Effectively, it takes around 3-4 times the duration of the file to correct the automated transcript, if you include the time for replays. It is also easy to lose focus on long files. So, remember to take breaks. Without the automated transcript, you may have to spend 8-10 times the duration of the file.

Of course, if you do not have the time, our transcribers will be happy to make the corrections for you. We guarantee 99% accuracy for our manual transcripts. Please do try it out.

Improved Automated Transcripts

Our latest speech and language models have been released. There are several new features in this release. The following is a list:

Acoustic Model: This is our fourth acoustic model trained on our data. The dataset contained mostly accented speakers (eg. Indian, African, Irish etc.). It also contained some noisy files. The accuracy of the automated transcript on accented files should be better now.

Language Model: We have added more data to our language model and doubled its size. The model now model has now been trained on around 46 million lines and has improved the WER by around 2%.

Punctuations: The biggest feature of this release is expanded punctuations. We now support all types of punctuations including quotes and hyphens. To our knowledge, nobody else including Google Web Speech, AWS Transcribe and Speechmatics supports quotes.

Speaker Turns: We also have updated our speaker turns model. The accuracy of the model is around 80% on long paragraphs. The automated transcripts will be better segmented now. We are currently working on adding speaker diarization to the automated transcript and it should be out soon. We do speaker turns a bit differently and do not require the number of speakers as an input. That is also one of our unique features. Google Web Speech does not support multi-speaker files and AWS Transcribe and Speechmatics require the number of speakers as an input for diarization.

This release also fixes the issue of missing predictions where some words, especially near speaker turns were not being transcribed. The automated transcripts should now capture all utterances, except filler words. We also benchmarked our model with LibriSpeech Clean and our internal dataset. The following are our numbers.

Dataset Type WER CER
LibriSpeech Clean Read speech 14.53% 5.85%
Scribie Internal Conversational 16.33% 8.82%

For comparison, PaddlePaddle numbers are the following:

Dataset Type WER CER
LibriSpeech Clean Read speech 5.4% 1.9%
Scribie Internal Conversational 28.34% 17.71%

As you can see, for conversational audio, our models outperform PaddlePaddle by a wide margin. We are working on improving our models for non-conversational audio as well. Our ASR is a DeepSpeech-based system and therefore a comparison with PaddlePaddle is a good benchmark for us. The Continual Learning blog post has some more details on how we trained our DeepSpeech models.

The automated transcripts are free currently, so try it out today!

Continual Learning for Speech-to-Text

Flawless transcripts and fast turnaround time are the hallmarks of Scribie. Not only are our transcripts highly accurate, but also priced reasonably. But have you ever wondered what makes that possible? The answer lies in constantly improving our speech-to-text engine, which assists our transcribers. We provide automatic word completion to our transcribers, and the better those autocompletes are, the less they have to type.

Our speech recognition engine is a Deep Learning system. For the uninitiated, Deep Learning is a subdomain of Machine Learning. It makes use of Artificial Neural Networks that, in a way, mimic the structure and function of the human brain. Our speech recognition engine is based on the DeepSpeech 2 network from Baidu, and written in PyTorch.

Scribie has a large dataset of audio and transcripts — over 100,000 hours at the last count. Training Deep Learning models over such a large dataset is very expensive in practice, as it requires a large number of GPUs and SSDs. For comparison, Baidu trained their models with 256 GPUs on custom hardware when they developed the DeepSpeech architecture. We don’t have the time or money to do that. So we developed an approach which we call Continual Learning.

Continual Learning

We first built and trained a large model with a 3,000-hour dataset. That took around three weeks on our rig. Since then, every month we have built a ‘corrections dataset’ of around 1,000 hours. This corrections dataset is made up of predictions from the previous model that were wrong, and then manually corrected by our transcribers. In each iteration we remove an equal amount of data from the previous training set and fine-tune the model over the newly combined data. This ensures that our model keeps improving over time.

Results

We have completed three such iterations and the results are promising. We have been able to consistently decrease the Word Error Rate, a common metric for automated transcription accuracy. The following is the chart of our WER.

We are providing free automated transcripts for a limited time, so please don’t hesitate to try out our online speech recognition system soon! Please note that we support only English at the moment and it works best for files with North American speakers and clean audio.

Deep Learning and AI has been in the news a lot lately, and there are concerns that Machine Learning will end up taking our jobs and replace humans. We have taken a different approach and built a system to assist our transcribers instead. Eventually, we want to reach a point where a human would have to spend just 10 minutes on a one-hour file, and still produce a highly accurate transcript of it. We still have a long way to go and we are working hard at it!

Did Trump Say I or I’d?

The Jury is out on what President Trump said: I or I’d. But what does the AI think? We put our free automated transcripts service to test on the following clip.

And here’s the result.

But with that being said, a president has been extremely generous with what he said. I like him a lot. I have a great relationship with them, as you know, have a great relationship with prime minister abe in japan, and I probably have a very good relationship with m gun f not care. I have relationships with people to surprise.

So our AI agrees with WSJ. President Trump did say ‘I’. So there you go!

The transcript is missing few words towards the end and we are working to fix it. However, if you have a clean audio file then head here to get a free automated transcript!