Silicon Valley might be the first place you think of when it comes to cutting-edge digital tech, but eyes are increasingly turning east, specifically to Tel Aviv in Israel. All of the major players in e-commerce have offices here: Google, Amazon and Booking.com.
The reasons for Tel Aviv’s rise to prominence are a combination of geographic, economic and cultural factors. The arid land has necessitated innovation in water technology, AgroTech and Smart Manufacturing, while the government places great emphasis on intellectual capital. It boasts an increasingly strong economy, low unemployment and a renowned academic system that fosters a talented, entrepreneurial workforce.
“Israel is called a start-up nation because it has the largest number of start-ups per capita in the world,” says Noa Barbiro, a native Tel Avivian and group product manager at Booking.com’s R&D Machine Learning Centre. “Being immersed in this collaborative community has huge benefits. I believe some of the future in tech is being created here in Tel Aviv.”
Booking.com’s office in Tel Aviv is an innovative tech lab home to small, collaborative teams of data scientists, developers, designers, copywriters and product managers that are busy experimenting and testing in a wide range of fields, such as machine learning, speech recognition and computer vision. They are tasked with looking for ways to make the customer experience even more accessible, personalised and streamlined.
[
]
The world of machine learning
At the heart of this future is machine learning. In its various forms, it’s helping to provide increasing personalisation to tailor to customer needs. But what is AI-led personalisation? As Booking.com senior product manager Shalev Barazani explains, “We’re building a destination recommendation engine that will help customers choose the next place to visit. And when they’ve made their choice and are searching for accommodation, we then present the most relevant, suitable options to reduce the amount of time they have to spend searching.”
We want customers to book with confidence at a property that perfectly suits their needs. Machine learning enables us to do this on a global scale
To give this idea a little added context, the global travel platform has millions of rooms available around the world, in all sorts of accommodation types. In a capital city like Rome, for example, they have over 11,000 options. That’s a lot for even the most ardent traveller to search through. But, thanks to a rich dataset and very accurate machine learning models, Booking.com can make finding the perfect place to stay much simpler.
To do this, they take a customer’s known behaviours, like habitually choosing city centre hotels with a gym at the lower end of the price spectrum, input this data into an algorithm (or model) and then present the most relevant options to suit that traveller’s preferences. Shalev sums up the value of personalisation through machine learning: “We want customers to book with confidence at a property that perfectly suits their needs. Machine learning enables us to do this on a global scale.”
Voice recognition
It’s estimated that by 2020 as many as 50 per cent of searches on the Internet will be done by voice. But voice recognition comes with its own unique set of challenges.
First, the language model has to translate the sound waves into what are known as “phonemes” – the individual sounds that make up words (take “cat”, for example – it involves a “k” sound, an “a” sound and a “t” sound in order to make the word). This is no easy task and requires a lot of data, not to mention considerable human input, to train an accurate model because clear rules have to be established. For example, someone saying “hmm” could mean they’re hesitating, agreeing or expressing their doubt. It’s essential, therefore, to train your language model to understand these nuances. And this is exactly the challenge the Booking.com voice recognition team is overcoming in Tel Aviv.
Booking.com's Booking Assistant uses natural language processing to quickly identify and respond to about 60 per cent of customer queries automatically
Another challenge being faced is what is known as “sentiment analysis”, which combines both the types of words that are being used and the tone of voice. In this way, voice recognition models can determine whether someone is angry, frustrated or happy. These are important distinctions when interacting with customers.
One of the upshots of such research is the provision of more powerful chat products. At Booking.com, its chat product, called the Booking Assistant, uses natural language processing to quickly identify and respond to about 60 per cent of customer queries automatically. These include things like: “What time is check-in?”, “Is parking available?”, “Is there Wi-Fi in my room?” and so on. This leaves customer service staff free to deal with more complicated enquiries that require nuance and empathy.
Computer vision
Computer vision uses maching learning to understand visual content by interpreting different assortments of pixels, which are combined together to form shapes. How does this benefit a business like Booking.com? Well, it makes checking thousands of accommodation images for graphic or offensive content much easier. It also helps improve the user experience for the customer by tagging and surfacing the most relevant images first. Say a customer is looking for breakfast or food options in a hotel, AI can identify the most relevant image to present to them.
The key at Booking.com is being able to test, test and test again
One of the great advantages in developing this type of tech in Tel Aviv is that there are other companies with cutting-edge knowledge and know-how to collaborate with. “We build on top of known and available technologies, but then encompass it with additional coding and algorithms that we build here in order to understand those things,” says Noa. The key at Booking.com is being able to test, test and test again, which ultimately leads to quick fails and much faster learning.
The way forward
The experimental, collaborative spirit of Tel Aviv is best summed up by Laurent Hallermeier, director of the Tel Aviv R&D Machine Learning Centre. “What makes Tel Aviv a tech hub and a leader in machine learning is a mixture of culture and the mature technology ecosystem that’s all around us. The Booking.com office here benefits from the Israeli entrepreneurial spirit and the work-life balance of a European company. While we share Booking.com’s core values, we’ve also created some complimentary, additional ones of our own. We’re collaborative, driven, positive, open and we like to have fun.”