The Quality annotations app allows you to request manual annotations for batches of outbound  / long form translations previously completed. During annotation, professional linguists go analyze the translation for errors and categorize them according to type and severity. You can later check their results in order to understand how quality is faring in terms of MQM score and most frequent errors.

Contrary to the MQM scores displayed in the Quality Reports app, which are obtained by our state-of-the-art AI Quality Estimation technology, all scores in the Quality Annotations app are the result of human annotation.

To access the app, click on Quality Annotations on the app panel to the left.

The app will contain two tabs, one for Requests and another for Routines.

What are Requests?

Requests are one time actions where the user can define the amount of translations to be analyzed, their corresponding time period and their origin in order to generate a report covering overall quality in CUA, MQ, while identifying errors and their severity.

What are Routines?

If you want to determine recurring quality assessment, it is possible to establish routines, which have the same outcome as request but are set to be repeated through time with a set frequency and defined timeline.

Creating a Request

Click the New Request button to create a new one-time request. The button may be placed differently if it's your first request.

There are different option when it comes creating a request:

Search translations

You can search and select from translations meeting certain criteria and create your annotation batch which will include a determined number of translations. To do this you can fill in the fields from left to right, top to bottom. New fields are unlocked as you set previous required criteria.

1 - Select the number of translations to annotate or a percentage of the total number of translations meeting the criteria.

2 - Select the time period comprising the translation activity.

3 - Select the pipeline group used by the translations you are monitoring.

4 - Select the specific pipeline (language pair) to analyze.

5 - Choose a specific application/integration or choose to consider all.

6 - Choose the translation step to analyze. Choose between MT (first step if your pipeline includes human translation) or the Final version before delivery.

7 - Set the word range. Only translations which source text is counting up to that number of words will be considered.

8 - Click Search. It may take up to a few minutes to load all eligible translation.