Salvatore Giammarresi

Airbnb’s Head of Localization

Interview by CAMERON RASMUSSON

Last year, Airbnb announced it partnered with Italian LSP Translated to create Translation Engine, a language solution for its massive collection of listings and reviews. Covering 60 languages in total, the sophisticated machine-learning process — a custom solution based on Translated’s ModernMT technology — seamlessly translates content for everyone from vacationers to perpetually traveling “digital nomads.” By some estimations, it’s the largest contract awarded in terms of total words processed.

MultiLingual reached out to Airbnb’s Head of Localization Salvatore (Salvo) Giammarresi to discuss his professional background, the experience of such a massive undertaking, and the state of the industry in 2022.

Advertisement

ML: Tell us a little about your own professional background. What was the career path that led you to Airbnb?

SG: I moved to San Francisco in the mid-90s during the first Dot Com era with a Masters degree in Languages (specialized in Linguistics) and a self-taught keen interest in AI and computer science. I discovered the world of localization during one of my earlier roles as Engineering Program Manager at Electronics for Imaging. I fell in love with it and realized that I wanted to focus my career in this field. I particularly loved how localization combined my passion for people, languages, cultures, engineering, products, business, quality, human-machine communication, AI, and operations. From the start it didn’t feel like just “work,” but instead it became my “passion.” As a bilingual speaker and son of emigrants, I could deeply empathize with those who were facing linguistic and cultural barriers in their life and in their work. I could also understand what was holding back many companies as they struggled to efficiently make their products locally relevant at scale. My professional career path then continued as Senior Localization Manager at Kana Software; then Director of International Product Management at Move.com; Vice President of Products at Classified Ventures; Senior Director of Localization Engineering at Yahoo; and Head of Content & Globalization at PayPal, before joining Airbnb as Head of Localization.

In parallel, my academic career led me to a PhD in Applied Computational Linguistics and becoming a visiting professor, teaching localization, at the University of Palermo in Italy.

ML: What were some of the factors that prompted you to accept your position with Airbnb?

SG: I was drawn to Airbnb because I felt that it had a unique and interesting localization puzzle to solve. In most products, the assumption is that your user is always in the same “locale” (mix of language and country). At Airbnb, by definition our users, particularly those traveling cross-border, were on the move going through two or more locales. In many cases, there might also be a linguistic and cultural mismatch between host and guest. I saw a clear affinity between Airbnb’s community and localization’s own mission of connecting people through cultural empathy and language.

I also loved that, while Airbnb was already localized in around 30 languages, there was still room to improve. I saw a lot of potential for innovation, and doing things differently. Indeed, since I joined, we have been pushing the boundaries and, in the process, helping Airbnb’s global community to connect.

The focus of the first year was the team itself and moving to a vendor model. This allowed us to scale and grow our operational bandwidth. We expanded the size and role of the internationalization engineering team. We expanded and refocused the localization program management team. We refined the role and scope of our language managers. Last but not least, we onboarded Translated as our sole human translation partner across all Airbnb content — which allowed us to automate many processes, and free up valuable time across our team, allowing for more and better involvement with our internal stakeholders.

The second year we focused on language expansion, and in record time, we doubled Airbnb’s supported languages and created a more robust global gateway.

Most recently we focused on improving the quality and UX of translating user generated content and this led to the creation of Translation Engine.

ML: What is Airbnb doing right now in the world of localization that you think offers a lesson to your industry peers?

SG: Our team operates under two foundational themes. The first one is to serve Airbnb’s global community of hosts and guests, partnering with all internal teams and functions to localize and internationalize everything that Airbnb creates, at scale, at a high quality, on time, and within budget.

The other foundational theme — from day one and everyday since — was to question everything and try to innovate wherever possible.

Our first innovation was to select and focus on just one language service provider, Translated. While the traditional wisdom is that companies should hedge their bets and use more than one provider, in my experience, this was causing more issues than it was solving. With one provider, we could work as one team and become true partners with common goals and metrics.

Another thing we did was to align our team with the values and business goals of the company. This has allowed our team, over time, to expand and deepen our role, and get involved further upstream in all processes.

Localization is viewed as a foundational asset at Airbnb. This was exemplified by Airbnb CEO Brian Chesky’s Nov. 9 product announcement and demo of Translation Engine.

Lastly, Airbnb is setting a new bar for the whole industry in regards to translation of user generated content (UGC). For many years Airbnb, and the entire industry, thought that “state of the art” was to display UGC in whatever language it was created in, and then offer a “translate” button that calls a third-party machine translation provider. In many ways UGC was deemed less relevant, and therefore it didn’t deserve the same high quality human translations.

We were inspired by this Charles Eames quote: “The role of the designer is that of a very good, thoughtful host anticipating the needs of his guests.” Translation Engine makes Airbnb an even better host to our community of hosts and guests, by presenting all content, whether it’s created by Airbnb or UGC, in the user’s preferred language.

ML: When did Airbnb first begin offering automated translation? What was the scale? What were some limitations?

SG: Airbnb has been offering machine translation by a single provider of its user generated content for many years. Some of the limitations have always been quality and cost. Using a generic engine meant that Airbnb couldn’t independently improve the quality. Cost started to become a factor as our UGC started to grow exponentially in line with the global growth of Airbnb.

ML: How has language access progressed on Airbnb since then?

SG: The main progress, in terms of language access, occurred in 2019 when Airbnb doubled the number of supported languages, bringing the total to 62. This ambitious language expansion program was one of the single largest ones in the localization industry, given its size and scope, and allowed more than one billion people to access Airbnb in their native language.

Advertisement

ML: When it comes to automated translation, do you have any stories or examples of the system working not quite as intended?

SG: When we launched Translation Engine, we gave the example of a seaside listing in Brittany, France: When guests clicked to translate the listing from French to English, the listing’s translated title read as “House feet in the water.” We knew that we could do better than that! Now, with Translation Engine, that same listing’s title will be translated to “beachfront house,” and that translation will happen automatically.

One of the reasons why Translation Engine bears its name is because — like an engine — it is made of many components, and the overall success of the engine depends on how all these components work together. An important part of the work was to find the right balance between quality, performance, and capacity. At the end we were able to improve the quality of our listings by 99%, while also improving the speed and the capacity.

ML: Let’s dig a little deeper into Translation Engine. I’m curious to know how it was developed.

SG: We worked with our best in class partner — Translated, leveraging their ModernMT technology — to create Translation Engine, a brand-new machine learning model and solution. Translation Engine is trained on Airbnb’s content to better understand the content it’s translating and continuously learns Airbnb’s and our users’ language and expressions. The end result is a higher quality, more human-like translation rather than an automated, one-size-fits-all computerized translation.

ML: How long has this translation technology been in the making?

SG: I first had the idea in early 2018, while I was interviewing for the Head of Localization role at Airbnb. There was definitely a way to make it easier for hosts and guests to communicate and help create more human-like translations. In 2019/2020, it really became one of our main focus areas.

ML: What were some of the challenges you encountered along the way? How did you overcome them?

SG: The biggest — but also most rewarding — challenge was to align across more than 100 people and different teams across the globe. While our team had the vision, set the direction, and built the Translation Engine technology in partnership with Translated, we needed to partner with all the teams that managed product flows where user generated content was being ingested and displayed. It definitely had its challenges, but it’s so rewarding when you see the final product and hard work the team accomplished.

ML: Tell me a little about the practical scenarios this will enable. What stress points will this relieve for travelers seeking a smoother experience? How about the hosts?

SG: Frictionless communication is at the core of creating human connections. Previously, guests on Airbnb would experience a localized product excepting user generated content, which was displayed in whatever language we received it. To display this content in the user’s language a user had to click on a “translate” button, which would use an off-the-shelf software translation service. Thanks to Translation Engine, when you land on a page, all of its content will be pre-translated in a preferred language — saving guests time, creating a better user experience, and making it even easier to book. Hosts will now be able to write their listing in their own language and rest assured that anyone around the world can read it in their own language. And the best thing about Translation Engine is that it learns and adapts to new content. The more information that’s added to it, the better the quality.

ML: A growing trend I see among my own colleagues is the “digital nomad” — professionals who maintain a lifestyle of perpetual travel while working via the internet. I’d imagine Airbnb has a clearer, more data-supported view of this lifestyle, since they’re likely frequent customers. What have you observed about these types of people, and did you have them in mind when developing these new translation services?

SG: The world is undergoing a revolution in how we live and work. Technologies like Zoom make it possible to work from home. Airbnb makes it possible to work from any home. As a result, consumers are demanding a newfound flexibility. According to a recent five-country survey that we commissioned, more than one third of self-identified remote employees say they would rather quit their jobs than go back to work in-person full-time.

Now, people increasingly are living on Airbnb, and they are doing so around the world. In a dozen-plus destinations, at least 50% of recent nights were for long-term stays — from Bangkok and Phuket to Buenos Aires. And as borders begin to reopen again, guests are increasingly looking to travel further: Between early 2021 and now, the most acceleration in growth for guests has been for trips of more than 3,000 miles from home.

Translation Engine will certainly serve this new, emerging need among guests who are living on Airbnb far away from home.

ML: Are there any other types of guests and hosts who will particularly benefit from this and the other new services you’re introducing?

SG: As US borders reopen — the largest country to reopen to foreign travelers arriving by commercial flights — Translation Engine will be particularly useful to travelers ready to travel longer distances and cross borders once again. In fact cross-border travel has been increasing steadily throughout the year, from 20% of gross nights booked in the first quarter, to 27% in the second, to 33% in the third. Even with most international travel having been shut down during much of the pandemic, connection between people of different backgrounds on Airbnb and the desire to travel across borders never went away.

Now, with the largest inbound travel market in the world open once again, many consumers are eager to take part. During the week following the Oct. 15 announcement of the Nov. 8 reopening date, nights booked by foreign guests for stays starting Nov. 8 increased by 44% as hosts in more than 4,000 different destinations across the US will welcome inbound foreign guests.

ML: I’d imagine these services have the potential to only expand as time goes on. What kind of a future do you see being paved by these new capabilities?

SG: Right now, search pages, listing pages, and the first few reviews of a listing will be automatically translated — but we’ll be rolling Translation Engine out further across the service. We’re incredibly proud of Translation Engine, and we are looking forward to introducing higher quality, more human-like translations platform-wide.

ML: Is there anything else you want to highlight that we haven’t yet covered?

SG: Airbnb is built on the idea of hosting. More than 4 million Airbnb hosts have welcomed more than 1 billion guests into their homes worldwide. Airbnb helps establish human connections through hosting. Translation Engine is our way of being great hosts to the millions of people who are part of the Airbnb community. Translation Engine unlocks all the user generated content, created by hosts and guests on Airbnb, making it easier for human connections to occur on our platform.

ML: Given the scope of the project, you must need some time to relax and unwind. What are some of the hobbies and interests that keep you occupied outside work?

SG: When I am not working on Airbnb projects, I spend time reading, which fulfills the lifelong learner in me. I love taking walks in nature and playing tennis. When time permits, I love advising companies and coaching people.

ML: When you look to the future, what most excites you about localization? What are the challenges and pitfalls that will need to be overcome or avoided?

SG: The internet has made the world smaller. Any company has the potential of reaching millions of people online. This is a clear trend that has been growing over time. According to Byte Level Research, every day about 1 million people join the internet for the first time, and only nine out of 100 new internet users speak English. Not only is English no longer the most-used language online, but the largest and fastest growing segment of users is made of all the “other” non-top-20 languages combined.

Within this context, the role of localization professionals is going to be even more critical. This is what excites me and this is what I tell people who are considering joining localization teams.

The opportunities I see for localization professionals on the buyer side is to have a larger impact in their respective companies, by connecting dots that no one else sees. Connecting people, growing the business and making companies more locally relevant at scale and efficiently.

For language service providers on the vendor side, the opportunity is to find novel ways to partner with localization buyers, to bring about process, engineering, automation, and operational innovations that will unlock even more value for all.

The challenges are that the role of localization as a function, on average, has not fully matured yet, and therefore it is not always properly leveraged within companies. There are very few functions and people in a company who have the depth and breadth of seeing how all the moving parts of a company coalesce into a product or the launch of a new service or campaign. Localization professionals must embrace this, while deepening, expanding, and leveraging their unique combination of cultural, operational, linguistic, engineering, and design know-how.

Opening doors: Airbnb.org’s effort to shelter refugees

The recent events in Ukraine provide a stark reminder to always be prepared for refugee relief. No one can foresee the turns of world events, and war, drought, famine, or natural disasters can create a refugee crisis before anyone has time to respond.

While the Russia-Ukraine war brought refugee issues to the forefront of the collective consciousness, it’s worth remembering that similar crises happen whether the world is paying attention or not. When refugees are in need of shelter, every open bed makes a difference. And there are few companies that know about matching people with beds better than Airbnb.

In 2012, an Airbnb host requested the company’s assistance in providing free shelter to individuals displaced by Hurricane Sandy. As Airbnb gained more experience providing shelter to those in need, the program evolved into a full-fledged nonpofit called Airbnb.org. Hosts can opt to provide rooms for free to those impacted by emergencies — they even earn a special profile badge for their community spirit.

“To date, Airbnb and Airbnb.org have connected more than 54,000 refugees and asylees — including from Syria, Venezuela, and Afghanistan — to temporary housing through Airbnb.org partners,” said company spokesperson Liz DeBold Fusco. “Last year, Airbnb.org announced the creation of its Refugee Fund and has galvanized more than 4,000 donors to further support its work with refugees and asylum seekers worldwide.”

To better serve refugee needs, Airbnb.org partners with International Rescue Committee, Church World Service, HIAS, CARE, and other local and global nonprofits. Good teamwork is key for an escalating crisis like Ukraine, where regional and international NGOs have a clearer view of the situation on the ground. The plan is to provide assistance for 100,000 Ukrainians.

“Airbnb is mobilizing hosts that want to offer their home for free or at a discount for this cause,” Fusco said. “Airbnb will also provide the services, technology, and other resources necessary to facilitate these stays via the Airbnb platform.”

“In addition to Airbnb.org’s efforts, we have been so humbled by the inspiring generosity of our community during this moment of crisis, with many booking stays in Ukraine with no intention of staying — simply to support local hosts,” she added. “On March 2 and March 3, there were more than 61,000 nights booked in Ukraine around the world, including more than 34,000 nights booked by US guests. Airbnb is temporarily waiving guest and host fees on bookings in Ukraine at this time, and on March 2-3, total gross booking value to Ukraine was nearly $2 million.”

With so much energy and willpower being directed toward helping Ukrainians, it’s a good thing Airbnb has built up significant experience in refugee aid. In August, for instance, Airbnb announced it would provide free temporary housing to 20,000 Afghan refugees — a goal the company met in February by assisting 21,300 individuals.

“[We] set a new goal of providing free, temporary housing to another 20,000 refugees from Afghanistan, Africa, the Middle East, Central and South America, and other regions,” Fusco said. “Airbnb.org’s offer to provide housing to up to 100,000 refugees fleeing Ukraine is in addition to this broader effort.”

According to Fusco, the thanks goes to so many individuals and homeowners who are eager to help out in the midst of an international catastrophe. That kind of motivation, taken collectively, makes all the difference.

“Anyone interested in getting involved with Airbnb.org should go to airbnb.org/help-ukraine, and support Airbnb.org’s initiative to provide housing to refugees fleeing Ukraine, by becoming a host or donating,” she said. “You don’t have to be an existing Airbnb host to participate — anyone can share their home with refugees through Airbnb.org.

Advertisement

RELATED ARTICLES

Subscribe Now for Full Access

Select a print or digital plan to get early access to MultiLingual and never miss an article.