Localization World is one of the few conferences attended by representatives of major clients of localization services.
Here you could find, for example, heads of localization departments in Oracle, Microsoft, Paypal, King.com, Blizzard, NetApp, GetYour Guide, ESET, Booking.com and other global brands. The 34th LocWorld, held in Barcelona from 14-16 June, was attended by more than 600 delegates, at least a quarter of whom represented client companies.
This time I had the good fortune to work on the conference's programming committee and act as a moderator for some of the talks. Here are some of the pieces of news about the international localization market which I was able to pick up during the event.
The bot was demonstrated by Wei Zhang during the innovation contest. To the translator, the bot looks like a Skype contact (or other user of a chat service). As materials arrive the bot issues new phrases that need translating. The translators start work on these phrases and may send the translation in the form of an answer. The result is a kind of dialogue between person and machine, with the option to type in commands. Instead of using a computer, you can take part in these dialogues using a telephone: standing in a supermarket queue on the Metro for example.
The bot is designed to allow real time translations within large groups of people- not necessarily professionals- who are cooperating on projects where time and scale are more important than a high level of quality. For example, it is able to support emergency and rescue teams, including volunteers, as well as crowdsourcing projects and Continuous Delivery projects. Microsoft is planning to launch the bot technology on an open source basis this June.
This course will be available on the Udacity platform, free of charge. It will have 5 hours of video, plus written texts and audio materials.
In the course Google staff will explain the basic terms and concepts, explain what skills are required and demonstrate the tools used in the localization process. The translation memory function will be demonstrated using the Google Translator’s Toolkit.
This introductory course will take one day and it will give those attending it a head start in the profession. And it is likely that the course will have thousands of subscribers- this is Google, after all!
Adaptive machine translation (AMT) technology caused a stir in the profession last year. This technology was popularized by Lilt, a Silicon Valley startup. AMT involves a machine translation service that is able to adapt itself to the author's terminology and style while it is working on the text. The program automatically substitutes phrases in the text and the user can confirm the substitutions by repeatedly pressing Enter and occasionally editing the translation. This enables the translator to work faster and faster- up to 3 pages an hour or more.
Example of how Lilt works
After Lilt's AMT technology, which is rather expensive for the Russian market, SDL Trados 2017 came out, and it is now becoming available as part of an increasing number of programs. For example, the Japanese translation platform YarakuZen, which was first presented in LocWorld. It is likely that one more European platform will be rolled out in June. In a couple of years, perhaps, most translators will be using AMT. It is a more organic process than just choosing repeated terms suggested by the translation memory. .
Digitizing and measuring absolutely everything is now a major trend in all areas of business. In the localization business in the last two years companies have started to measure and analyze datа on their productivity and translation expenses. Forward-thinking companies are paying more attention to how localization can improve their sales, bring them more traffic and improve User Experience. And they are changing their policies accordingly.
In the Data-driven round table, representatives of the Starwood hotel group and Booking.com presented complex case studies. There are too many hotel pages on their sites to make it economically viable to translate between all the possible language combinations by hand. Booking.com has 1.3 million hotels and operates in 43 languages. They therefore need to decide which hotel pages need to be translated into which languages.
Starwood has therefore created a formula to deal with this issue – it looks at the income that each hotel gets from visitors of particular nationalities. By using the formula to select spending priorities the company's localization service was able to save $45 million in expenses at the lowest possible cost.
Booking.com analyzes its internet traffic and chooses a budget for translating a hotel page once the number of page views from a browser based in a given country exceeds a certain threshold. For example, it makes sense to translate pages for hotels in New York into Japanese, but as yet there is no point in doing this for hotels in the Republic of Georgia.
Electronic Arts's localization department monitors players' behavior when playing computer games. For example, as the company's audio translation manager, Mario Bergantiños explained, if half Russian gamers skip through the loading credits on the FIFA football simulator, then there is no need to provide an audio translation for them- subtitles will be enough.
Takeaways from the round table:
Collect data, and check that it is accurate.
Use analytics (for example, GA or Yandex Metrika) to select the languages into which it makes sense to translate the site..
Carry out A/B testing of the localized pages.
Provide the company's management with figures for the income generated by investing in localization, rather than its cost.
Other related news:
Skyscanner's research shows that client satisfaction depends on how quickly the client is answered, in his own language, in the company's live chat. This is an example of a situation where translation speed is more important than quality.
King.com has calculated that 86% of non-English language advertising campaigns outperform English-language ones. Takeaway: it is essential to translate advertising materials.
Major translation services try to measure the quality of their translations in points and display this information in graphical form, with categories for components, languages and translators.
This is generally done as part of the final stage, in which 5-100% of the finished translation is checked: the proofreader marks any mistakes in style and terminology and rates them (how serious or minor they are). The result is an objective quality rating, on a scale of, for example, 1 to 100. The manager of the translation department can monitor these ratings on his display panel or 'dashboard'. When a company works with more than 20 languages and translation volumes running into the hundreds of thousands of words, such a dashboard can be an indispensable tool for visualizing a vast amount of data. With this quality model it is easy to experiment with machine translation or new suppliers and then review the results of such experiments.
In general such a system is an irreplaceable asset for the head of a translation department. But such a system requires the introduction of an LQA system: first the accepted materials are selectively checked and then this process is automatized.
In the forum “Quality at the Speed of Now — It’s All About Customer Experience!” heads of localization departments from Intuit, GetYourGuide, LinkedIn and eBay said that they are currently using Excel, and their own internal software, for this purpose. But they are certainly interested in the specialist software solutions that are available on the market. .
There are a number of editing programs that have already been on the market for a couple of years, but they are very little used, with maybe a few tens or at most a couple of hundred users. These include LexiQA, globalReview (Kaleidoscope), ContentQuo, Change Tracker, TAUS Quality Dashboard etc. Traditional translation memory platforms are also gradually adding new functions that allow editors to highlight spelling errors etc. and rate the translations on a scale. As far as I know, memoQ и XTM have gone furthest in this direction.
An ideal, specialized editing tool should:
All this sounds complicated, but the available tools are moving in the right direction. The question is who will create the simplest and easiest tool to work with, and be able to sell it to the largest number of professionals.
It is important for those translation companies that wish to focus on high-level clients keep up with these tendencies. On paper the audience in LocWorld looks very promising for representatives from translation companies, but in practice, it is very difficult to get accepted by such large corporations.
These clients are contacted by dozens of sales managers and they are used to brushing them off without hesitation. Negotiations in which the client demands 'The same level of service as our current providers, but 10% cheaper' typically end almost as soon as they start. Managers of translation departments, who have been able to get budget approval to attend an expensive conference (typically $1,500-3,500 per person, including conference entry and other expenses) have normally long ago covered their clients' demands in relation to language services. In order to push away their current providers from attractive contracts, the new vendor must offer something quite new, and interest the client in a unique skill or project.
That is why these conversations always start with discussing innovations, new languages, trends in the industry, lingering unresolved problems, and experimental pilot projects. But after the translation bureau has shown some interest in the major brand's latest models, the parties may then go on to talk about traditional matters such as providing large volumes of translation services for perhaps a couple of million dollars.
PS: one more interesting thing from the conference: the bicrawler program, which can create thematic translation memories in a number of languages, from other peoples' websites.