Top 5 reasons affecting quality in automated transcription

by May 6, 2019Automated Transcription

Quality of automated transcription is improving all the time as the accuracy increases. Therefore, increasing amount of people are able to select automated transcription service instead of traditional human-made transcription.

Especially in the English-speaking world automated transcription has become a real and reliable option when in need for converting audio files to text format.

Several reasons affect the quality of automated transcription. In this post we will go through the five most common reasons affecting the quality, as well as tips how you are able to increase the quality by yourself!


1. Audio language


A relatively basic factor affecting the quality of automated transcription is the audio language. Currently the best quality is achieved with audio files that are in English (including its various accents) within almost all of the available automated transcription services.

If the language itself is common worldwide, the quality will also be higher compared to languages which do not have that many speakers in the world. This is due the fact that speech recognition technology is based on machine learning.

The technologies learn faster in case the language is spoken globally and it is used in different services. The service providers are also able to modify their technology to support that specific language.

For the time being, the only way to guarantee the highest quality is to record your audio in English or in some other common language, e.g. Spanish.


2. Speaking accent


Accent of the speech should be as clear as possible. Most automated transcription service providers have made it possible to select the language accent.

For example, if the discussion is in British English, there is a greater likelihood to reach high quality transcript if you use speech engine specifically designed for the accent.

To increase transcript quality, it is recommended to select a service provider that offers that specific accent.


3. Over talk on recording


Over talk is one factor affecting the quality of automated transcription. Speech recognition technology may not be able to identify the speakers in case they are talking simultaneously.

When recording the audio e.g. while having an interview, you can focus on to avoid over talk and encourage others to do the same. When people take turns while speaking, the quality of automated transcription will be much higher. Of course people talking over each other is a natural and sometimes unavoidable feature of a discussion, but it should be avoided as much as possible.


4. Background noise


Background noise interferes the speech recognition and naturally decreases the quality of automated transcription. Background noise may be caused by other people talking in the background, or by traffic, music, or windy weather.

When choosing the place for the interview or discussion, make sure that the environment is as quiet as possible. Therefore, a public café may not be the best option since there are other people causing background noise. In the best environment your discussion can be recorded clearly without echo or extra noises.


5. Talking speed


In the top 5 of reasons affecting the automated transcription quality is the talking speed. Slower talk equals better quality.

We have noticed in practice that speaking slowly, the transcript quality is higher compared to fast-speed talking. When you speak at a slow pace, you also articulate clearly, which again results in higher quality in speech recognition.

It may not be easy to speak slowly, especially if you are enthusiastic about the topic under discussion. However, focusing on the slow speed, you will end up with high quality automated transcript.