Experts predict that by 2020, up to 50% of all searches across the internet will be voice searches. Sales of devices that allow voice-based search are increasing at an amazing rate, with Google and Amazon selling over 10 million such devices in 2018.
Multilingual Voice Search
The way we search for things through voice-based devices is very different from when we type keywords into a search bar. Even when speaking to a machine, we want human conversation. For brands who are working globally, this means not just translating their websites into other languages or even localising their content according to the target country but doing it in a way that imitates local speech. We all know that the way we write is different from the way we phrase things when we speak. That is why it is essential to translate voice search not only as a translator but also as an interpreter, anticipating how someone would search for a product when speaking to a device.
How Voice Search Translation Is Done
Voice search engines are trained to recognise context and produce the most natural responses. These responses will be different in another language, or even in the same language but a different country. Word-for-word translation that sounds unnatural is never a good idea, and it becomes even more important in voice search translation to trans-create rather than to translate too literally. Voice search translation requires a specific set of translator skills, including localisation, trans-creation, and, in particular, an intrinsic knowledge of how things are said in everyday conversation.
In addition, the way we speak and ask questions changes over time. New words and expressions constantly appear in languages, while other phraseology becomes less used or outdated. Even in the same local area, language varies from one social group or community to another. Translators who work with voice search have to stay as current as possible and be aware of these variations.
The text to translate is taken from speech and needs to be translated into natural speech, too. Translators also record an audio of the translation which is then used to train the machine. This allows search engines to collect as many local voice variations as possible.
Translating Featured Snippets
Most voice search results come from a website’s featured snippet. A snippet is the answer box of information that comes up at the very top of the first page when you perform a screen-based search. The featured snippet is pulled from Google’s first page of organic results and is full of specific information that answers the question. This is what makes it so useful for voice search in particular because voice search is most often performed by asking a question.
Creating succinct, informative content that answers specific questions and can be pulled into snippets is essential if you want Google to pick your website for its voice search results. The best snippet-quality content provides answers to long-tail keyword searches, meaning searches that are very specific to your product or service. This type of content includes lists, tables or paragraphs explaining specific points of a topic that your customers are interested in. It is a good idea to have a combination of simple keyword and long-tail keyword content on your website.
If you operate globally, you have to translate any content that can be picked up as featured snippets into all relevant languages, making sure that the translations are fully localised. Furthermore, those snippets have to sound natural when spoken out loud since they will be read out by a voice search device such as Google Home or Amazon Echo. In an increasingly technology-based world, it becomes all the more important to translate as a human in order to engage with global audiences.
By Maia Nikitina
