Understanding Engine Failovers

Failovers are a safety feature in the Smart Model Selection process. It means that the primary engine selected is replaced by the next engine in line in the priority list. While rare, it may happen for various reasons, most notably:

• Translation Quality Estimate (QE) failover
• Timeout failover
• Availability failover

The purpose of these mechanisms is twofold: The TQE failover helps improve translation quality, while the timeout and availability failovers aim to reduce the latency of responses as much as possible.

Translation Quality Estimate failover

For Translation QE, the system assesses the quality score of the translation. If it falls below the defined threshold:

1. If Quality Failover is enabled, the system automatically generates a second translation using an alternative machine translation engine. The second translation is scored and the score is compared to that of the original translation. The translation that has the higher score is delivered.
2. If Human Translation is enabled and available in your Language IO application, the agent receives a prompt to either send the content for Human Translation or to proceed with the translation that was delivered. The Human review step is entirely optional.

For more information, see How does Language IO act on Translation Quality Estimate (TQE) scores.

Timeout and Availability failovers

Timeout

When using automated translation services that require calls to translation engines, Language IO enforces a standard timeout. This means that if the engine takes longer than the set time limit to return a translation (for example, 5 seconds), the system fails over to the next vendor in the priority list to attempt to complete the translation.

Availability

The Availability failover works in a very similar way as a the Timeout. The only difference is that instead of the request being received but timing out, the event that triggers the failover is the detection that the engine is not available to process the request in the first place (for example, because the server is down and the call returned a 500 error). Since there is no answer, the process fails over to the next vendor in the priority list. The process is automatic.