Choosing and tailoring machine translation engines demands skillful integration with current frameworks. Reports by leading consulting firms show that numerous innovation efforts fail because of inadequate integration with existing systems. At Intento, we acknowledge this issue and work to overcome it.
Security and machine translation providers
Picking a machine translation provider requires thoroughly evaluating their security measures and certifications. Check if they have ISO and other pertinent certifications. Assess their security protocols, procedures, and whether they perform penetration tests for system security. Remember that government standards vary from industry standards, so ensure the provider adheres to applicable regulations. Selecting a provider with a robust security framework helps safeguard your organization’s data and intellectual property.
Data Privacy and Compliance Considerations
Safeguarding data privacy is vital when partnering with machine translation providers, particularly with regulations like GDPR and CCPA. Establish standard contractual clauses to protect data privacy. For instance, GDPR clauses must be in place for lawful data export when transferring data from Europe to the US. Addressing privacy concerns is achievable by implementing appropriate measures within your organization’s structure.
Data Protection and Retention
When partnering with machine translation providers, ensure your data is protected at all times, both in transit and at rest. Verify that your vendor uses proper encryption and limits data access at every location.
Consider data retention policies as well. Some providers may not clarify how long they will retain your data, potentially storing it indefinitely. Also, some may use your data to train models for other clients by default. To prevent this, actively opt out of such policies for each system.
Lastly, examine how your data will be used to train engines for other clients. Make sure your vendor’s policies meet your data privacy and confidentiality needs. These steps will ensure data protection and compliance with relevant laws and regulations.
Data Locality
Consider data locality when choosing a machine translation provider. Certain countries, like China and Brazil, enforce strict data sovereignty regulations, preventing specific data types from crossing borders. When doing business globally, it is essential to understand data flow and select providers that adhere to local data sovereignty rules. This may require using different providers or provider regions for various countries. Compliance with each country’s regulations is crucial not only for efficiency but also to prevent legal issues.
Legal and procurement
Involve legal and procurement teams in selecting machine translation engines for compliance with regulations and contracts. Software-as-a-service and AI purchases demand a thorough terms-of-service review, including sanctions and restrictions related to company locations.
Localization departments may be experienced in buying human services, hosted software, and licensed software but could miss key data processing aspects. To satisfy contractual requirements and regulations, legal and procurement teams should review machine translation providers’ terms of service for compliance with data privacy laws and data processing obligations.
Additionally, these teams should address liability and indemnification concerns in case of data breaches or security incidents. Evaluate the chosen technology solution’s scalability, flexibility, and any potential limitations or risks tied to the provider’s business model or tech stack.
Financial and corporate
Besides budget allocation, evaluate the return on investment (ROI) for machine translation technology implementation. Assess the technology’s benefits, including increased efficiency, enhanced productivity, improved translation quality, and long-term cost reduction.
Also, consider the expenses associated with training and retraining employees to use the new technology and the cost of ongoing maintenance and updates.
Costs vs. ROI
Balancing costs and ROI can be challenging for localization departments. While you may incur costs, the benefits of implementing machine translation engines might surface elsewhere. For example, deploying machine translation in customer service can lead to a 20-fold reduction in labor costs by supporting non-English speaking regions with lower labor expenses, even with a 10-fold increase in MT costs. In such cases, beneficiaries might report to different executives who enjoy the “free” ROI, leaving the localization manager to handle the costs. This situation demands careful consideration.
Another concern is the discrepancy between costs and ROI. If costs are the only visible factor after a year of production and the ROI is not, it can cause issues within the company. Therefore, consistently evaluate and measure the cost-effectiveness of implementing machine translation engines.
Business workflows
Ensure business workflows align with human workflows, as modifying the latter can be even more challenging than integrating software with APIs.
LSPs and outsourced services
Maintaining quality control and data security can be challenging when outsourcing to LSPs. Establish clear expectations and communication channels to guarantee data privacy and quality standards.
Select an LSP with machine translation experience and the ability to customize and optimize MT engines for your needs. A skilled LSP can provide insights into workflow optimization, integration with existing tools, and translation quality. However, set clear guidelines for collaboration and communication to ensure a successful partnership.
Raw MT workflows
For successful raw machine translation implementation, prioritize user adoption by offering training and support to employees, support agents, and legal departments. This ensures they can effectively use the system in production.
End-user involvement
For an improved experience, include end-users in technology selection and evaluation. Give them a voice in decision-making and let them witness the technology in action. At Intento, we prioritize human evaluation as part of our process, believing in human involvement over sole reliance on automatic scores. This approach ensures that our technology meets end-users needs and expectations.
Collecting data
Gathering performance data like post-editing content is vital, but retrieving it from some TMS can be difficult, so plan for data collection. To enhance user experience, allow end-users to give feedback on the technology. A feedback form can offer valuable insights into their experiences. Start collecting feedback early, before the system is in production for a year, providing ample time for adjustments.
Existing software systems
When choosing machine translation engines, you have to integrate existing software systems. Luckily, there are several ways to achieve this. One option is using high-level API connectors for seamless integration. Also, most software systems provide API connectivity, and many TMS feature XLIFF APIs for offline translations, making integration feasible.
Integration with Machine Translation
Integrating machine translation into a TMS can be done in several ways. Plugins enable MT integration into existing systems, with options varying based on the system. A benefit of using third-party plugins is that the vendor usually handles security checks, reducing the burden on your security team, who might otherwise need to review 400 responses on a security questionnaire.
You can use built-in MT systems within the TMS or use MT connectors from certain providers. However, building and supporting individual connectors can be costly if you use multiple MT engines across various internal systems. It is often better to use existing software for integration.
Consider the integration type needed for system feedback. MT plugins within the TMS will not provide post-editing information, so an alternative integration, like an XLIFF interface or another API, is necessary. Additionally, gather usage statistics to understand consumption across departments and projects. This may require integrating with the MT provider’s Usage API or usage consoles. Having statistics in your Power BI or Tableau is ideal, but it calls for different integrations to work seamlessly.
Existing business relations
In complex translation scenarios like audiovisual translations, you may work with multiple vendors, including language source providers, quality assurance firms, and video production companies. Each vendor contributes their software, contracts, and expands your software footprint.
Introducing new technology
When exploring new technology, you might find appealing tools that seem promising. However, adopting them might necessitate switching to a different language service provider, even if you have worked with your current one for a long time. Thoroughly evaluate such decisions before taking the plunge to ensure the benefits are worth the change.
Renegotiating with service providers
When partnering with outsourced services, evaluate how the advantages of new technology will be distributed. Both the enterprise and service provider must be willing to adapt their work to ensure a fair allocation of costs or faster results at the same price without compromising quality.
When renegotiating contracts with existing service providers and introducing new technology, it is crucial to have data points supporting your case for adjusting translation payments or securing better rates. For instance, you could implement new technology while adhering to previous agreements for six months. During this time, gather data points from your TMS system. Analyzing this data can offer insights into the new technology’s performance and help you make a more informed decision when renegotiating your partnership with service providers.
MT Improvements
Ensure that every dollar spent on post-editing becomes a dollar spent on improving machine translation. As technology advances and investments are made, you should see an enhancement in machine translation quality. Continually renegotiating contracts with service providers every time a new MT model is implemented can be frustrating. Ideally, you and your service providers should collaborate so that as your MT engine improves, your return on investment increases. Unfortunately, some providers charge a flat rate for post-editing, regardless of the MT engine quality. Altering this arrangement will take effort, but maximizing your ROI in machine translation is necessary. Remember, every journey starts with a single step.
Takeaways
- Carefully consider existing frameworks when integrating machine translation technology to prevent implementation failure.
- Choose a machine translation provider by evaluating their security, certifications, data privacy, compliance, protection, retention, locality, and legal and procurement aspects.
- Involve legal and procurement teams in decision-making to ensure regulatory and contractual compliance.
- Balance costs and ROI when selecting and customizing machine translation engines.
- Align business and human workflows, and establish clear communication channels and expectations when outsourcing to LSPs.