Hybrid Translation for Websites

Introduction

In the previous two sections, we explored the respective advantages and shortcomings of human translation and machine translation. At the conclusion of both chapters, we mentioned the capability of using a combination of human and machine translation to:

• Apply Neural Machine Translation for high-volume, low-traffic areas
• Use human translation for brand sensitive or high-traffic areas, such as homepages and landing pages

In this brief section, we’ll dig a little deeper into the value of applying this hybrid translation approach to certain business use cases. You’ll learn about:

Explicit Use Cases for Human Translation: Some industries and businesses require absolute linguistic accuracy, crafted by human linguists. You’ll learn which organizations benefit most from this human-first approach.

When Machine Translation is Best: You’ll get the inside track on what kinds of website content can be accurately translated by NMT platforms, to save you time and money.

Tips for Leveraging the Hybrid Approach: We also provide actionable advice for identifying what sections of your website—or what types of content—benefit most from a human + NMT translation approach.

The Differences Between Human and Machine Translation for Websites

This chart provides a solid overview of what we’ve covered so far regarding human and machine translation, and how a hybrid approach can reduce time to market and translation costs:

A Closer Look: Use Cases for Human Website Translation

The human + machine hybrid translation approach is great for cost-conscious organization because it minimizes the amount of human translation applied to a localization project.

However, we recommend using human translation for specific industries and use cases. In these scenarios, the need for absolute linguistic accuracy, cultural nuance and persuasive communication can only be guaranteed by using human linguists:

• Medical, healthcare, health science
• Technical industries, such as chemical manufacturing
• Financial sectors
• Luxury products and other high-end sectors (where brand voice is paramount)
• Other highly technical content (technical specifications, etc)
• Legal content
• Website cart/conversion paths (where accuracy is important) and “My Account”-like user areas

A Closer Look: Use Cases for Machine Website Translation

The product titles in your inventory are important for several reasons, including SEO. Human translation of product titles ensures accuracy, nuance, and great SEO results. But for repetitive content—say, phrases like “100% cotton” or “assembly required”— machine translation can easily and quickly handle the work, resulting in time and cost-savings.

Here are a few other instances in which machine translation is an appropriate choice:

• Straightforward, non-technical general content (such as blogs)
• Information on products with low price points or short lifecycles
• Product specifications
• Charts and tables
• Descriptions of support, maintenance or warranty programs
• Functionality such as forms and survey instruments

Saving money with machine translation can free up additional funds for high-quality human translation, which is critical for high visibility sections of your site.

Using Human-Machine Hybrid Translation for Websites

Firstly, work with a website translation vendor that understands your business goals and budget and can properly define the scope of your project. Great vendors help you determine what areas of your site may benefit from human translation, and which can be easily handled through machine translation.

Use Your Analytics

You can help make those decisions by examining your website’s analytics. Some questions to consider:

• Which webpages, products or services your multilingual customers are most interested in?
• Where does their interest wane?
• How deeply do they dive into specific website sections or product categories before bouncing out?

Leverage human translation for the high interest content. You may be able to use machine translation for all other content. Be sure to continually review those analytics and adapt to customer behavior as it evolves over time.

Translation Quality Varies by Language

A quick word of caution: While Neural Machine Translation is leagues better than previous approaches to machine translation, the accuracy of some NMT languages are more advanced than others.

• Generally speaking, NMT systems do a good job of translating content into widely spoken Western languages such as Spanish and French
• However, complex non-Western languages such as Arabic and Chinese are presently too nuanced to be well represented by NMT translations

It’s possible to “train” NMT software to become better at translating in specific “target” languages, this this requires time, effort and resources.

Read more about website translation in our ultimate guide to website translation.