How To Get Better Accuracy with Automatic Transcription (2019)
Save money with automatic transcription by following these best practices when recording and creating your audio.
Automatic transcription is the most cost-efficient way of transforming your audio into text, costing less than 1/10 of manual solutions, and being done instantly without it costing extra for urgency.
However, the one point where automatic solutions struggle to compete is with accuracy, which can vary from around 95% to less than 10%.
The good news is that if you are responsible for creating the audio file to be transcribed, there is a lot that you can do to ensure that the automatic transcription comes back close to perfect! Keeping these best practices in mind is very beneficial, as it will help you save money in the long run:
Save your audio in WAV format
There are over 30 types of audio formats used throughout the internet. Although you’d have to be a professional to know most of them, I’m sure you’re familiar with the most common ones, such as .MP3 or WAV. That being said, there is only one audio format you need to remember if your goal is to get the best results possible: .WAV. These files are raw and uncompressed. This means that there was no loss of information while transferring these audio files into your computer. While most audio formats sound the same to an untrained ear, the file format makes a huge difference for automatic transcription algorithms. By using the lossless .WAV format, you can ensure that the audio quality is the best possible in order to maximise your transcription accuracy.
The easiest way to save your audio using the .WAV format is by using the free, open-source mixing software Audacity. Here is our guide on how to save your audio with the .WAV format using Audacity.
Remove Background Noise
Background noise is one of the main culprits for a low accuracy rate with automatic transcription. Since the algorithm tries to understand everything that is being said in the audio, any background noise can affect the final result of your transcription. There are many ways to remove background noise from your audio file and make the audio crystal-clear, both for human listeners and for software. The easiest way, of course, is to record in a room with good acoustics and as little distraction as possible. However, this is often easier said than done, especially for recordings such as meetings and lectures. If you have already recorded your audio file and want to remove background noise, you can do so using the free software Audacity. Here is our own a guide on how to remove background noise using Audacity.
Enunciate Clearly
This point might be a no-brainer for most people working in audio, but it’s easy to fall back into normal speech patterns when conducting a face-to-face interview or presenting in a meeting. In fact, there are so many things to pay attention to in these situations, that unless you’ve trained yourself to enunciate clearly, it’s likely that this concern is at the back of your mind. Fortunately, there are many ways to train yourself to do it subconsciously, and produce not only better quality audio, but greatly improve your presentation skills! We don’t have a guide for this one, as we would rather leave this to professionals. That being said, there are plenty of articles online that can help you train yourself to enunciate more clearly.
Let us know if there are any other resources you found to be helpful with this!
Use quality microphones
Your recording microphone makes a huge difference on the audio quality and automatic transcription accuracy. Naturally the more expensive microphones, used by professional recording studios, will guarantee the best quality and make audio easiest to understand even for human listeners. However, you don’t need to shell out a lot to get a decent solution for your audio needs. We recommend the Professional Condenser Microphone Venoro Broadcasting for those just starting out with audio recording. At $15, it is an excellent budget solution for microphone quality issues. Given that increasing your transcription accuracy saves you transcription time that can be used in other tasks, this microphone will pay for itself after a few transcriptions!
These are our recommended best practices to improve your transcriptions. However, you don’t need to implement all the measures in this list to get ideal results for your audio files. Even at 50% accuracy, automatic transcription can save you hours of transcription work in the long run!