If you are using Smartling's AI Toolkit, the AI Toolkit Dashboard allows you to monitor its benefits and better understand how Smartling's AI-powered features help optimize your translation processes.
Accessing the AI Toolkit Dashboard
The AI Toolkit Dashboard can be accessed from the AI Hub:
- From the top navigation of your Smartling dashboard, click into the AI Hub.
- Click into the tab AI Toolkit Dashboard.
From this dashboard, you can get an overview of your usage of Smartling's AI Toolkit, and check how its features are benefitting your translation process.
Tip: To enable Smartling's AI Toolkit or to learn more, please reach out to your Smartling Customer Success Manager.
Word Usage
The Word Usage widget in the AI Toolkit Dashboard helps you keep track of your Smartling capacity bundles and usage.
In addition to your Smartling platform subscription, you can purchase a pre-defined number of MT words and/or AI Toolkit words. These capacity bundles help keep track of how much content gets translated in your Smartling account - either with Machine and LLM Translation, or with one of our AI Toolkit features.
These packages can be purchased separately based on your translation needs, and can be used during the duration of your Smartling platform subscription.
The Word Usage widget shows how much capacity has been used up for each of these bundles. You can check both the percentage and the actual number of words used.
Info: A warning is displayed once the word usage capacity for your account is running low or has been reached. We recommend reaching out to your Customer Success Manager to discuss your options to keep translating content in Smartling.
Tip: Paid-off overages may still be included in the usage numbers displayed on this page. Any discrepancies will be addressed in a future update. If you require any clarifications, please reach out to your Customer Success Manager.
MT Words Used
This widget shows the number of words that were translated with an MT provider, by using either a Machine Translation workflow or one of Smartling’s account-wide MT integrations (including Smartling Translate, Smartling's Zendesk Support Plugin, site-wide machine translation for the GDN, dynamic translation with the ServiceNow Connector and translations using Smartling’s Machine Translation API).
MT suggestions in the CAT Tool, strings using TM Match Insertion, as well as translations using Smartling’s AI-Powered Human Translation (AIHT) workflow do not count towards your MT words capacity.
AI Toolkit Words Used
This widget shows the number of words that went through a translation step where at least one feature of Smartling’s AI Toolkit is enabled.
Translations using Smartling’s AI-Powered Human Translation workflow (AIHT) or other workflows managed by Smartling Language Services are not counted towards your AI Toolkit usage.
AI Toolkit Dashboard reports
The AI Toolkit Dashboard provides pre-configured reports that allow you to check how Smartling's AI Toolkit helps optimize your machine or LLM translation process.
The reports can be filtered by date range, language pair and workflows.
Info: The data displayed in the AI Toolkit Dashboard is refreshed 24 hours after the translations have been submitted from a Post-Edit, Edit or Review step.
Translation Memory Leverage Reports
Smartling allows you to enable TM Match Insertion for Machine Translation. This means that entries from your Translation Memory can be used instead of a Machine Translation if a match above a specified percentage is available. In addition to this feature, AI Adaptive Translation Memory can be used to optimize available matches and thus increase the TM leverage for Machine Translation workflows.
The reports in your AI Toolkit Dashboard allow you to check how these two features enhance your translation process. The graphics display the number of words from your Translation Memory that were leveraged for all of your MTPE workflows (i.e. workflows using Machine Translation, followed by a step for human Post-Edit, Edit or Review).
To qualify as "discounted leverage" in the context of these reports, the fuzzy match percentage must reach at least the first discount tier as per your Fuzzy Match Settings. For example, when using Smartling's default Fuzzy Match Profile, only a fuzzy match percentage of at least 85% is considered as leverage in the context of this report.
Workflows with AI features enabled in the translation step, but without a human Post-Edit, Edit, or Review step, are not included in TM Leverage reports. Words that move through these workflows do not count towards discounted leverage, as discounted leverage only occurs when a human is involved in a Post-Edit, Edit, or Review step and the fuzzy match profile is applied to that step.
The following reports are available to track TM Leverage and AI Adaptive Translation Memory in MTPE workflows:
1. TM Leverage in MTPE Workflows
This report shows the percentage of words going through an MTPE workflow within the last 30 days that benefitted from either:
- Discounted leverage without AI Adaptive Translation Memory,
- Discounted leverage gained by AI Adaptive Translation Memory or
-
SmartMatch
The white, unlabeled portion of the circle graph represents words that were machine translated and not TM-leveraged.
2. AI Adaptive Translation Memory performance
This report allows you to compare the Translation Memory leverage without AI Adaptive Translation Memory to the leverage gained by using AI Adaptive Translation Memory within the last 30 days.
Only content going through an MTPE workflow is taken into consideration for this report.
3. Translation Memory Leverage in MTPE Workflows over time
This report shows month by month how many words going through an MTPE workflow have benefitted from discounted Translation Memory leverage. By hovering over the timeline, you can see the percentage of translated content that leveraged your Translation Memory in one of the following ways:
- Discounted leverage without AI Adaptive Translation Memory,
- Discounted leverage gained by AI Adaptive Translation Memory or
- SmartMatch
Click "View settings" to access your MT Profiles, where you can enable TM Match Insertion in Machine Translations.
Click "View report" to open the full report, which can be downloaded in a CSV format.
Tip: To achieve cost savings for the Post-Edit step based on fuzzy matches, ensure to apply your Fuzzy Match Settings also to the post-translation step.
Language Quality Estimation (LQE) Reports
As part of Smartling's AI Toolkit, the Language Quality Estimation (LQE) Agent uses AI to predict the quality of each string translated via machine translation, giving insight into how much editing may be required by a human editor before a machine translated string can be published.
1. Language Quality Estimation levels
This report displays the Language Quality Estimation level for all machine translated words that were processed by Smartling's Language Quality Estimation feature within the last 30 days. By hovering over the graphic, you can see both the percentage and the word count for each level.
2. Language Quality Estimation & human edit rate
This report displays the total number of machine translated words which were processed by Smartling's Language Quality Estimation feature within the last 30 days.
You can see a breakdown of this total number by Language Quality Estimation level, both in terms of word count and percentage.
This report also displays the Average Edit Rate and Median Edit Rate.
Average Edit Rate: This metric shows the average number of edits made to translations by human editors. It’s calculated by taking the total number of edits and dividing it by the total number of translated strings.
Median Edit Rate: This metric measures the amount of editing required for translated content, normalized to account for the length of strings. It represents the middle value of the edit rates when they are sorted in order, providing a more robust measure of the typical amount of editing needed. The normalization process ensures that extremely long or short strings don’t disproportionately affect the measure.
Click "View settings" to learn how Language Quality Estimation can be enabled for your MTPE workflows.
Click "View report" to open the full report, which can be downloaded in a CSV format.