Overview
The Scribe AI system is designed to assist MT companies in better managing the quality assurance portion of their work flow and reduce unnecessary review of documents by QA, while ensuring that poor quality does not slip through the typical random audits of QA personnel.
How it Works – Calculating a Score
Basically the system “reads” every document created by MT’s on the Scribe platform. It reads them for the following information:
| 1. | Spelling |
| 2. | Grammar |
| 3. | Blanks – denoted by (__ |
| 4. | Watchlist – words on a specific list that Scribe maintains and used by only selected Cobrands |
For all cobrands spelling and grammar are weight 25% each and blanks 50% to come up with an overall score. If cobrands use the watchlist this is also included in the weighting. The score is then further weighted by the number of errors versus the size of the document. The more errors in a smaller document the lower the score.
The system takes all these measures and comes up with a final score on a scale of 0 to 100. This is the “AI Score” that the system shows to people and uses to determine what to do with documents.
How it Works – Processing Documents
Based on the score that is calculated and the AI “Thresholds” that are set for groups and users the AI system determines what to do with the documents. There are four thresholds that can be set:
| 1. | MT low |
| 2. | MT high |
| 3. | MD low |
| 4. | MD high |
The combination of the thresholds and the scores process documents according to the diagram on the next page. The thresholds are set by looking at the individual MT Group to set MT low/high and the MD Group (or MDUser which overrides MD Group) on the Admin platform.
