In Smartling, you can select your preferred translation option for each content type and language pair by configuring multiple workflows.
Each workflow has a Translation step, where a translation is applied to your content. You can choose between Human Translation, Machine Translation (MT) and AI-Powered Machine Translation. Additionally, all translation options can be combined with optional workflow steps for human validation.
Each translation option has its advantages. To achieve the best possible results, it is important to match each content type with the appropriate translation workflow, and consider the value and exposure of the content you want to translate.
Let's explore some of the most common translation workflow options, along with the pros, cons, and use cases for each.
It is important to remember that workflows without a human in the loop will miss the benefits of visual context or translation instructions, which help ensure contextual accuracy.
Machine Translation (MT)
Machine translation has come a long way since its beginnings. Nowadays, neural machine translation (NMT) adapts to grammatical structures and allows for a more accurate and natural sounding translation of entire sentences.
However, not all MT engines produce quality results for all language pairs and content types. It is essential to test an MT engine's translation with a specific language pair and content type to assess the quality against your standards.
In Smartling, MT can be used as a stand-alone translation solution, or it can be combined with additional steps for human editing ("MTPE").
MT Only (Raw MT)
Smartling integrates with multiple Neural Machine Translation engines and Large Language Models and even offers an intelligent routing service which automatically picks the best MT provider for your content and language pairs (Smartling Auto Select).
✅ Pros: Raw machine translation is the fastest and cheapest translation solution. Using machine translation for the right content types can help you optimize your translation cost and speed.
❎ Cons: Without the assistance of Smartling's AI features or human post-editing, it is difficult to maintain your unique brand voice and the translations don't always sound idiomatic and native to your audience. Raw machine translation without any human post-editing is typically not saved to your Translation Memory, and therefore cannot be leveraged in future translations. In the long run, trying to cut costs by avoiding a Post-Edit (MTPE) step could end up costing more.
⚙️ Use cases: Machine translation on its own is ideal for testing purposes or for translating low-visibility content with a short shelf life. It is suitable for translating user-generated content like support tickets, forums or customer reviews, product information catalogs, blogs, or low-liability instructional content like training materials or subtitles. Raw machine translation is not an appropriate choice for high-visibility creative content, such as brand slogans.
Tip: Smartling offers custom MT engine training, which can help significantly enhance the machine translation output. Trained MT engines can be used as a stand-alone translation solution, or as part of an MTPE workflow.
Machine Translation with Post-Editing (MTPE)
Smartling supports the combination of machine translation with proofreading by a human linguist. This is achieved by adding a Post-Edit step to a machine translation workflow.
This type of workflow is usually referred to as MTPE, or Machine Translation Post-Editing. It is a cost-saving option that delivers higher-quality translations compared to raw machine translations.
✅ Pros: The human editor can tweak translations to adhere to your Style Guide, ensure Glossary entries are applied correctly, and make light edits to correct any grammatical or cultural mistakes in the translations. Combined with a human editing step, machine translation can produce good results for content that is not highly creative but requires a literal, straightforward translation.
❎ Cons: Although this is another cost-effective option, editing can delay the turnaround time for translations. If used without Smartling's AI features, MT translation quality is not comparable to human translation.
⚙️ Use cases: Like raw MT, MTPE is not an appropriate choice for high-visibility creative content, such as brand slogans. However, with the right editor in the workflow, these translations can work well for technical content, such as user manuals, knowledge bases, medical and legal documents.
Tip: MTPE workflows can be combined with Smartling's AI Toolkit to benefit from innovative AI features and achieve a translation output that is tailored to your preferences.
AI-Powered Machine Translation
Generative AI has revolutionized the world of localization. Smartling's innovative AI features allow you to achieve a high-quality translation output that is tailored to your preferences at a fraction of the cost of traditional human translation or MTPE approaches.
You can either use Smartling's AI Toolkit in combination with your regular MTPE workflow, or take advantage of Smartling's fully pre-configured AI-Powered Human Translation (AIHT) or AI Translation (AIT) workflows.
AI-Powered Human Translation (AIHT)
As an alternative to human translation, Smartling's AI-Powered Human Translation (AIHT) combines on-brand translations at human-level quality with improved speed and cost savings. This is a workflow offered exclusively by Smartling Language Services and backed by Smartling’s Satisfaction Guarantee.
For each string, the best available machine translation is selected by AI from a range of NMT providers and Translation Memory. This starting translation is enriched by AI-powered features like AI-Enhanced Glossary Term Insertion, AI Fuzzy Match Repair and AI-powered post-editing to clean up formatting and whitespaces. The translation is then reviewed by a professional human linguist to ensure it reaches human-level quality.
✅ Pros: This workflow is fully pre-configured for best results and produces human-level translation quality at half the cost and twice as fast. Translations using an AIHT workflow are saved to your Translation Memory, which can then be leveraged to reduce the cost and turnaround time for future translations.
❎ Cons: For highly creative content (e.g. slogans), a transcreation workflow may be required. If very rapid turnaround is needed for time-sensitive and low-visibility content that does not require human review, MT might be a suitable alternative.
⚙️ Use cases: AI-Powered Human Translation is ideal for marketing websites and other high visibility web content, application content, product manuals, subtitles, medical or pharmaceutical content, user facing instructional content, packaging, or binding legal documents.
AI Translation (AIT)
Smartling's AI Translation (AIT) is a fully pre-configured translation workflow that offers quality comparable to MTPE at a much reduced cost, with near immediate ROI for high-resource languages. This is a workflow offered and managed exclusively by Smartling Language Services.
AIT uses all of the same AI features as our AI-Powered Human Translation workflow, but it does not include any human review. For a limited group of lower-resource languages a light-touch human post-edit step is required.
Tip: Your Smartling representative will be able to advise whether your target locales require a Light Post-Edit step or if our fully automated AIT solution is sufficient to produce high-quality translations. Please note that the AIT workflow is only offered for language pairs with an English source locale.
Translation turnaround times are based on API processing speeds, with most translation requests delivered near instantly for target locales where no Light Post-Edit is needed. For target locales where post-edit is required, Smartling can process a standard minimum of 5k words per day.
AIT best fits into an overall solution mix that also includes workflows with human translation or review, like AIHT or Professional Human Translation. Translations using other workflows with a human in the loop are saved to your Translation Memory.* Therefore, when used as part of a solution mix with human-in-the-loop workflows, the AIT workflow is able to continuously leverage the benefits of human validation from other workflows via the Translation Memory. This ensures that the AIT’s style and terminology does not drift from any fixed pre-existing translations, reducing the need for rework while maintaining ROI.
* See this article for full details.
✅ Pros: This workflow option is fully pre-configured for best results and provides translations at a quality level comparable to MTPE at a much reduced cost. As an automated translation solution, AIT offers faster turnaround times than human-in-the-loop solutions. For lower-resource target locales, a Light Post-Edit step is added to ensure a high-quality translation output.
❎ Cons: Since this workflow does not include any human validation or only light post-editing, it is less suitable for highly creative or high-visibility content. Additionally, files that are heavily formatted, that are otherwise complex, or that require DTP may not be good candidates for AIT as a human in the loop is usually required.
⚙️ Use cases: AIT is a great choice for translating lower visibility content, technical documentation, knowledge bases, SEO content, training data, internal instructional material, opinion portals or surveys, boilerplate legal, or anything time sensitive.
Tip: To enable this workflow for your account or to learn more, please reach out to your Customer Success Manager.
Human Translation
Smartling Language Services offers professional human translation services. Alternatively, content can be translated in Smartling by a translation vendor of your choice.
Professional human translation and editing
Most translation vendors include an Editing step in their translation offering. This means that as soon as a first linguist has translated your content, a second linguist proofreads the translations.
✅ Pros: The editing step ensures that the translations contain no mistakes and sound native to your target market. Your Style Guide and Glossary are taken into consideration by the linguists, so they can apply your preferred terminology and tone of voice. Human translations are always saved to your Translation Memory, which can then be leveraged to reduce the cost and turnaround time for future translations.
❎ Cons: There are no disadvantages in terms of translation quality. However, human translation takes longer to complete than machine translation. Depending on your chosen vendor, the quality of content and your needs, you may be faced with additional Project Management fees. While the initial costs are generally higher compared to machine translation workflows, careful human translation can help avoid costly and time-consuming re-translation at a later point.
⚙️ Use cases: A two-step workflow including professional translation and editing is a great choice for most content types, including high-visibility, creative content like marketing campaigns, collateral for print, or any culturally specific content.
Tip: Smartling's AI-Powered Human Translation workflow is a great alternative to classic human translation, offering the same level of quality at half the cost.
Review and sign-off by your internal team
Members of your organization who are native speakers of your target languages can be set up as internal reviewers in Smartling and approve, amend or reject the translations delivered by your translation vendor.
✅ Pros: Your internal team members know your product and brand best. They can tailor the translations to accurately reflect your brand image.
❎ Cons: As your internal teams typically can't fully dedicate their time to the proofreading process, an internal review step often leads to delays. A well-maintained Style Guide and Glossary can typically eliminate the need for it.
⚙️ Use cases: An internal review step can help ensure that your desired brand image is preserved, especially for content types that are highly visible and difficult to amend at a later point (such as print material). For critical legal documents, an internal review step allows your local legal teams to sign off on the translations.