There are several approaches to translating your customer service communications, but not all of them deliver the speed and high quality your customers expect. Find out what your customer experience needs are and why Unbabel stands out.
Unbabel
DeepL
Lionbridge
Lilt
Lokalise
Smartling
Language I/O
Unbabel
DeepL
Lionbridge
Lilt
Lokalise
Smartling
Language I/O
Unbabel
DeepL
Lionbridge
Lilt
Lokalise
Smartling
Language I/O
Unbabel
DeepL
Lionbridge
Lilt
Lokalise
Smartling
Language I/O
Unbabel
DeepL
Lionbridge
Lilt
Lokalise
Smartling
Language I/O
See How It Works, Schedule A Demo
See why customers love Unbabel
Machine translation-only solutions for customer service translation
Unbabel versus Google, DeepL, and Microsoft
What MT-only solutions do
Machine translation providers translate very broadly without human input. Models are trained on general data sets, meaning they do not offer industry-specific translations. There are often free versions available, APIs to connect to software systems, and tools to adapt the AI to your data.
Google trains its models using general data but does not focus on the customer service domain. With AutoML Google customers can adapt models to their data, however, Google does not offer services to evaluate, retrain or maintain the MT ecosystem. Domain expertise and resources would be required to self-serve. Google does offer customization through glossaries and APIs to connect to software systems.
DeepL’s sole business focus is machine translation. They offer 26 language pairs and the choice of tone for nine languages. They train their models using general data, not customer service data. DeepL does not currently offer tools or services for model adaptation to fine tune to specific customer use cases. Integrations are available for CAT (Computer-Assisted Translation) tools or via APIs, but integrations to CRM tools like Salesforce or Zendesk are currently not available.
Microsoft Translator is a pure-play machine translation system offering a variety of language translation applications. Microsoft’s machine translations use generic data that respond to a broad set of use cases. There aren’t any customer service data modules on offer to date. There are tools to adapt models to customer data, however, do not offer resources to evaluate, retrain and maintain the MT ecosystem. Engineers will need to connect the solutions to your current systems and workflows via APIs. Microsoft has experience in customer service use cases, but only for static (FAQ) use cases, not dynamic use cases such as a chat or email.
- Machine translation models optimized specifically on customer service data.
- Adaptation services to automatically evaluate and retrain models, supported by a dedicated team of domain experts.
- Human evaluation layer to improve translation — both in-message and over time.
Language Service Providers (LSPs) for customer service translation
Unbabel versus Lilt, Lionbridge, and RWS
What LSPs do
Language Service Providers (LSPs) primarily focus on localization. They leverage people, processes, and technology to deliver turnkey translation projects. They almost always include third-party machine translation in the workflow.
Lilt leverages a blend of professional translators, technology, and processes to deliver localization. They integrate into CRMs like Salesforce and Zendesk, have a suite of translation solutions like glossaries, termbases, and translation memories, and use machine translation to make suggestions to their translators. Their translators are fast, but not trained to manage immediate conversations in the customer service context. Lilt leverages external machine translation providers such as those listed above, trained on general data.
Lionbridge has years of experience delivering localization projects for a wide variety of use cases. They have a broad solution in which they mix machine translation and customize projects based on client needs. This means it might take time to set up and build out workflows specific to customer service. Machine translation use will help advance the workflow, but the ongoing learning is not fed into the underlying model and retrained.
RWS has a rich array of translation solutions, software products, and services to deliver localization: They might be the LSP with the broadest set of offerings across services and software. The approach and technology behind it are designed to optimize translators and their workflows and facilitate communication with clients. While they support customer service use cases, today they do not support dynamic communication channels like chat and email, nor do they leverage customer service-specific machine translation models.
- Translation workflows specifically built to meet speed, spikes, and fluctuations of customer service.
- Rapid setup and CRM integration to translate chat, email, and FAQ channel.
- Machine translation models optimized on customer service data.
Translation Management Systems (TMS) for customer service translation
Unbabel versus Smartling and RWS
What TMSs do
Translation Management Systems (TMS) are software to organize, monitor, assign and deliver translations, mostly in localization. Third parties can access to review and translate content.
Smartling makes managing translation projects simple with all the expected translation tools and easy access for third-party translators or reviewers. Smartling combines humans and technology and integrates into Zendesk and Salesforce, but does not have an offering to accommodate email and chat workflows for dynamic communications.
In addition to their LSP capabilities, RWS also has a TMS solution that helps businesses manage translation and localization projects alongside translators and reviewers. This makes RWS a fairly complete tool for localization work, but they currently lack the off-the-shelf workflows to support customer service communications, whether that’s email or chat, and it would be necessary to assign translations ad-hoc.
- Out-of-the-box customer service workflows for email, chat, and FAQ channels.
- Visualization, tracking, and reporting in the Unbabel Portal.
- Machine translation and human layer optimized for customer service needs.
Direct competitors for customer service translation
Unbabel versus Language I/O and Lokalise
What the direct competitors do
Other solutions apply third-party machine translation systems directly into customer service channels — like email and chat — with a tangential or optional human layer in the workflow.
Language I/O is built to deliver customer service translation at scale. They leverage third-party machine translation models from Google, DeepL, and Microsoft and select the optimum translation from any of the sources. Language I/O integrates into several CRM systems and has the option to escalate to a human for some, but not all, languages.
Lokalise’s core offering helps companies rapid-release localization, especially for websites, apps, and other user interfaces. Lokalise also translates customer service chat conversations integrated into Zendesk and Intercom. They use Google’s machine translation tool as the provider, and also can layer human reviews on top. They rely on Google’s ongoing improvement to their underlying algorithm as opposed to customizing models to the customer.
- A human quality assurance layer integrated into translation workflows.
- Quality feedback loop that improves underlying translation model.
- In-house machine translation models optimized on customer service data.