The aim of contributions is to calculate each agent's contribution to the final quality of the translation.

The task is considered to have reached its final quality when it has passed through all hands/steps and has arrived into Delivered. Agents can be linguistic tools or human contributors. 
Please be aware that contributions are calculated only once the task has been delivered.

To access the query, go into the Query Factory of your project, choose Translation batch and Contribution by timeframe and start filtering with your criterias.

The table shows three types of data, accessible in the "Columns" button :

- Volumes (matchVolume)

- Individual contributions (matchQM)

- Cumulative contributions (matchQMReached)

Here are the definitions of the categories of data and their applications : 

Individual contributions (matchQM): Each agent's contribution is counted individually. We want to know how much this or that agent has brought us closer to the final quality.

Cumulative contributions (matchQMReached) : The progression of quality of a task is calculated considering its progress through the workflow. We look at the state of the task at a given point, and count its difference to the final quality, taking into account all modifications made at previous stages. Previous individual contributions have been added up to give the cumulative contributions. 

Volumes (matchVolume): number of words in each category. Allows weighting and calculation of overall contribution per agent.

Now, let's take an example : 

For simplicity reasons, we are looking at one project, on one languages pair, with one contributor per human step.
Workflow in this project is Translation>Review>Validation>Delivered, with MT in Preproc.

For the example, we are focusing here on 2 categories : 100% matches, and new words (match under 49%). 

For 100% matches

We can see that for this category of words, of which there are 24, it has been first handled by the memory. Memory is responsible for 94% of the final quality. 
Then, it went into Translation. Here, the user modified some content and is responsible for 5.8% of the final quality. After translation, segments categorized as 100% matches have reached their final quality (matchQMReached100 indicates 1) . 

For new words

For this category of words, the first agent that will work is the machine translation (Deepl). 537 words are then going into Preproc and filled by MT. Deepl did a good job as it is responsible for 91% of the final quality. 

New words are then handled in Translation, where it brought us 7% closer to the final quality. 

Finally, in Review, the user brought the final touch (last 2%), and we reached the final version of the translation.