Scribie transcribers, are you looking to learn more about the difficulty level label? You’re in the right place.
As you know, we’re committed to delivering at least 99% accuracy, regardless of the audio file. Now, some files are easier to work with, while others can be tricky. In any case, we aim to consistently deliver our brand of quality. That’s why we have a method around this.
For us to know what we’re working with, we categorize each file according to their difficulty level. This helps us gauge the work required for each job.
Below, learn about how we handly varying file difficulty levels, allowing us to consistently deliver great transcripts and keep our operations smooth:
The Difficulty Labeling Method
The 3 Steps
Step 1: We generate an automated transcript. This will serve as the initial reference.
Step 2: We estimate the required work to get the job done.
Our machine learning algorithm assesses the automated transcript. Based on our dataset (i.e. previous transcription files), we estimate the number of words that are likely to be corrected by our human transcribers.
This estimate is highly-accurate (with about 5% margin of error) for 95% of our files.
Step 3: Finally, a file is categorized based on the difficulty level. Based on the estimation, a file will get one of these three labels:
- Low: <10% changes
- Medium: 10-20% changes
- High: >20% changes
To clarify, this difficulty level refers to the audio file’s quality rather than the transcript’s. The “% changes” refer to the number of words that are likely to the corrected the professional transcribers.
How these Indicators Work
As the name implies, these estimates serve as indicators. While no system is perfect, this allows us to operate with a consensus.
Each transcriber would have their own proficiency levels and preferences. Depending on the particular transcriber’s subjective assessment, a transcription job might appear easier or more difficult in comparison to another transcriber’s assessment.
At any rate, this common point of reference serves as a guideline for transcribers when gauging a file.
How This Affects the Four-Step Process
To add, as we go through the process, there will be fewer changes required. QCs will have to make fewer corrections compared to Proofreaders. This is the inevitable nature of quality assurance.
The Thought Process
We value you, our transcriptionists. As you know Scribie is a high accuracy transcription service that depends on your manual work.
Consider that these estimates are based on our historical experience. Our machine learning algorithm estimates based on the previous files that we’ve worked with.
Also, note that the incentives are aligned with each job’s required level of effort.
Got more questions or ideas?
Check out our FAQ page and engage with the community in our Slack channel.