COVID's impact on the travel industry last year was devastating. This year, however, leisure travel is bouncing back. As the vaccines roll out, a majority of people say they will travel this year, according to Travel Pulse. With many having carried over vacation days from last year, more than two-thirds of respondents say they'll take international trips ranging in length from seven to 14 days.
According to a recent study, about 90 percent of customers will research trips online, and 82 percent will book online. Another study by Common Sense Advisory indicates that these customers are much more likely to book on websites and apps that are in their native language, even if they speak English. If you want to capture your share of the travel bookings, translation of your website and apps will be critical this year.
Because 2020 was so challenging, many businesses in the hospitality industry are short on cash. Finding a lot of money to do the localization required may be a struggle. Machines are becoming much better at travel translations, however, and can translate much of your customer-facing content. The ability to gain quality machine translations (MT) means you may be able to complete the most important translations without blowing your budget.
The first MTs, such as Georgetown IBM, ALPAC, and Meteo, were rule-based. Human translators manually developed the rules. These machine translations were limited because of the large number of exceptions to the rules and the amount of human expertise required to derive the rules.
In the late 1980s, statistical machine translation (SMT) was born. SMT uses statistical models that learn to translate text from a source language to a target language. Because the method is data-driven, linguists do not need to develop the rules. SMT could translate much faster than rule-based MT. One drawback, however, is that it focuses narrowly on the phrases being translated rather than the context. It also ignores important syntax distinctions and requires careful tuning. While SMT is still appropriate for some types of translations, this technology has largely been replaced.
Neural machine translation (NMT) is one of the newest forms of MT. NMT is based on artificial intelligence and uses neural networks to produce high-quality translations. The machine is trained using text that humans have translated, then builds its artificial intelligence so that it can decode a text in ways similar to how a human would do so. NMT's strength is its ability to learn directly how to map from input text to output text.
Deep neural machine translation (DNMT) is an extension of NMT and uses multiple neural networks. DNMT can deliver real-time translations for chat applications and discussions and provide batch translations for documents.
Essentially DNMT works by encoding each word in the source text as a number until the whole sentence is decoded. Then, the numbers are input into a neural translation model, resulting in a different series of numbers. The new series of numbers are translated into the target language. The network is trained with millions of sentence pairs (such as English to Chinese or French to German) so that the neural network is refined.
The result is high-quality translations that will work, with a little human re-editing, for many aspects of your website.
Machine translations, which are less expensive than those done by human linguists, can be used on large portions of travel content. These include:
The remaining portions, including the user interface, apps, booking agents, and marketing and advertising materials, require experienced translators. By using MT when appropriate, you can translate your travel website without spending more than is necessary.
We understand the cost constraints caused by COVID-19. Andovar has ready-to-go trained travel translation engines specifically for tourism industry content. Book a meeting with us today to see how we can help you claim all those travel dollars.