What Now for Chatbots in Travel?

By Scott Crawford, Vice President of Product Management at Brand Expedia.

Originally appeared on Tnooz.com

2016 was the year when chatbots emerged as a new interface for consumer interaction.

Advances in artificial intelligence technologies – such as neural networking and natural language processing – have allowed brands such as Facebook, Google, Apple and Amazon to offer conversational products, letting consumers order products or map their journeys through speech or messaging.

As we enter 2017, chatbots will become commonplace, particularly in the travel industry. Machine learning will allow chatbots to become more and more sophisticated while customer expectations will rapidly evolve in tandem. On-demand 24-hour information and service will become commonplace.

When online travel agents democratised travel 20 years ago, they gave the power of flexibility and choice to the consumer. One could argue that the paradigm of booking a holiday online hasn’t evolved in huge leaps since then.

But chatbots represent the next seismic shift that will evolve not only the travel booking process but also the customer service experience for decades to come.

In the competitive world of travel, many in the sector are asking how they can take advantage of these new technologies. This year Expedia launched a Facebook chatbot, a new Expedia skill for Amazon Alexa and a chatbot for Skype.

As many travel brands face up to the challenges of launching their first bot, Expedia has been through that process already. Here are some of our observations.

Travel: the ultimate chatbot challenge?

There’s no escaping that booking travel is exceptionally complicated. It might not be immediately apparent how chatbots could make things simpler, given the layers of information which need to be gathering and collated in rder to make a travel purchase.

But there are plenty of areas in the travel booking process where chatbots can be implemented in order to improve the user experience.

Travel has an unusually long funnel compared with other ecommerce sectors. In the early stage, customers are unsure of the details of their trip, looking for flights or hotels in various destinations and time periods. Users further down the funnel might be business travellers or frequent fliers, who are more sure of what they’re looking for and need less guidance.

Users at each stage have very different information requirements, and so your chatbot needs to be designed to reflect multiple variables.

We see too many brands design a chatbot first and then think about its purpose to the customer afterwards.

First, think about bots in terms of solving a particular consumer problem or need.

In our experience, a chatbot interaction is more likely to result in a successful outcome for the customer in situations where the task the bot is resolving has clearly defined parameters. This means that chatbots, for the moment at least, don’t have much of a role at the top of the funnel in the research phase, as this is a highly subjective and open-ended experience for customers.

It’s difficult to design a single-purpose chatbot that will suit all consumer needs. That said, chatbot technology is getting more sophisticated by the day, and we expect to see lots of progress here over the next 12 months.

Know your limits

Chatbots are still a new technology, so it’s important to not overreach your design. Once you’ve isolated a specific stage of the funnel where a chatbot might be helpful, build a chatbot for that stage of the funnel.

Current chatbot technology is not advanced enough to allow for multipurpose bots: they need to be designed for single uses, such as collecting pieces of information from customers specifically to book a flight, or providing date ranges of available hotels matching certain criteria.

That said, new innovation is happening at a rapid pace.

Our recently launched Skype chatbot is the first bot experience on Skype to connect a traveler to a call with an agent within the platform. Customers can easily search for and make a hotel booking, or manage select elements of travel bookings, including hotel or flight confirmations or flight cancellations.

And if a traveller has an additional request that is not yet supported by the chatbot, Expedia will handoff the experience to an Expedia travel representative, or the traveller can call directly from within Skype for no charge.

For brands just starting to experiment with chatbot development, be clear and upfront about the purpose of your chatbot with customers. Define that first, provide your customers with some basic information about how to structure their queries, and a productive resolution to the problem is more likely.

As with any new technology, it is also important for the experience to feel as natural and intuitive to the user as possible.

Making the chatbot feel conversational is paramount, and small tweaks to how the bot communicates can result in big changes to customer behavior.
Looking to the future

There’s a vibrant and growing ecosystem of startups developing chatbots for all stages of the travel funnel. New companies such as Mezi, KimKim, and Pana have sprung up to help facilitate the booking process for flights, hotels and entertainment: once the customer has provided the initial information and search terms it becomes much easier for chatbot technology to help by automating key steps in the booking process.

For travel firms the key to capitalising on chatbots is to think small first: isolate pain-points in your booking funnel and ask yourself whether it could be resolved by a conversational function. It’s important to be honest here, as there’s a high price to pay for implementing a chatbot incorrectly.

Focus your chatbot on one or two functions at first, such as returning search results based on an input of a date period, and test and learn based on the data you gather.

The travel industry can build on what has already been achieved: there are already more than 11,000 chatbots on Facebook Messenger.

Additionally, continued mobile penetration provides the perfect context for chatbot growth.

At Expedia, we understand that you cannot constrain consumers to a particular platform; they will want to use the medium that best suits them. Many times this is across multiple devices, so the chatbot experience needs to be as sophisticated and easy-to-use as desktop.

As travel brands continue to experiment and release more bots into the ecosystem, insights will begin to show how consumers want to engage with chatbots – the better data and insights at our fingertips, our expertise will grow.

This is a viewpoint by Scott Crawford, Vice President of Product Management at Brand Expedia.

Image by Wutzkoh/BigStock.

Expedia’s Nautilus Travel Search Engine: Overview and Applications

By Thomas Crook, Manager of Data Science and Technical Product Management, Expedia Search Group, and Neelakantan Kartha, Senior Technical Product Manager, Expedia Nautilus Team

History and Introduction

Expedia has been researching natural language processing for more than five years and completed version 1 of the patent-pending Nautilus travel search engine in 2012. Nautilus is composed of an NLP language parser and a probabilistic travel entity selector we call “SmartFinder.” Client applications pass Nautilus a natural language query (e.g. “Hotels in Bellevue, “Beach Hotels”, “Hotels near Space Needle in Seattle”, “Las Vegas”) and Nautilus returns ranked lists of travel entities, such as hotels and regions, that are most relevant to the query.

In July of 2012 Expedia launched “Semantha,” an internal employee hotel booking site powered by Nautilus. Semantha accepted natural language queries and displayed a ranked list of hotel results on a map view of the destination most relevant to the user query. The Nautilus team used Semantha telemetry and qualitative feedback for three years to continually test and improve the Nautilus search engine and build confidence in its ability to successfully handle real world natural language travel queries before launching it for external use.

Nautilus and Expedia’s Search Anything

We started A/B testing a Nautilus-powered “Search Anything” tab on the Expedia.com US home page in September of 2015 to gauge customer feedback and search input trends, and have continued testing and adding incremental improvements to the feature ever since. For example, we were surprised by the number of people who were directly entering itinerary numbers so we added a feature that returns a link to users’ itineraries at the top of the search results when they enter an itinerary number.


Figure 1: We were surprised at the number of people who directly entered itinerary numbers into the Search Anything form

This past July we rolled out Search Anything to the majority of our English-language platform sites under the Expedia, Travelocity, Orbitz and Wotif brands that enables us to understand local context in search results. For example, searching for “Hotels in Newcastle” on Wotif.com will return hotels in Newcastle New South Wales in the first two positions, versus the same search on Expedia.co.uk which will surface hotels in Newcastle-upon-Tyne.


Figure 2: Localized Search Results

The most popular Search Anything query categories we have observed in 2016 include:

all inclusive
ground transportation
phone number
customer service
multiple destination vacation packages

Expedia is also testing Nautilus on other parts of its sites, beyond the Search Anything tab. In conjunction with our machine learning sorting algorithms, Nautilus is powering a hotel search result refinement test that allows users to surface hotels that meet their unique needs – such as “pet friendly hotels with a kitchen.”


Figure 3: Nautilus Search Refinement on Hotel Search Results

Nautilus Technical Overview

At a high level, here’s how Nautilus works:

The Nautilus orchestration service takes an input query and forwards it to the NLP service. The NLP service tags the query with named entities (e.g., Name, Location, Amenity, Other) and concepts (e.g., HOTEL_STRUCTURE, HOTEL_ATTRIBUTE, HOTEL_STAR_RATING, PRICE, RELATIVE_DATE). The tagged query is then passed on to the SmartFinder service, which is powered by machine learning models derived from various sources. The SmartFinder service uses the tagged query returned by the NLP service in combination with machine learning models to determine the lists of travel entities such as hotels and regions most relevant to the query. The Nautilus service constructs the final response using the outputs of the NLP and SmartFinder services and returns it the calling application.


Figure 4: Nautilus High Level Architecture

A Vision of Voice Disruption

By Tony Donohoe, Chief Technology Officer and Senior Vice President of Technology at Expedia Inc.

Simplicity is driving technology. Thirty years ago, technology was defined by a disc. Since then, there have been advancements just about every decade to bring us to where we are now – a mobile driven world, where you can search and book at any point, in any time, on any device. Recently, mobile has provided a huge shift across industries, driving 50% of traffic coming from mobile. But as technologies evolve, there’s a desire to have personal interactions through technology. Until recently, Voice User Interface was an area considered to be artificial intelligence, but now VUIs have become more commonplace, and people are taking advantage of the value that these hands-free, eyes-free interfaces provide in many situations.


Voice is potentially challenging for us in that there is no structured interface. We have trained Expedia users to click their mouse in a certain way or put their thumb in a certain field to get a response. We’ve trained them to declare where they’re going to and when. We’ve trained them to search for a single product at a time. That won’t be possible in voice. The consumer’s expectation will be, “did I get an answer to my unstructured question?” We need to build the capacity to handle unstructured voice search queries and produce structured search results. That’s a great challenge and opportunity, but we’re one of the few companies that will be able to do it.

At Expedia, our mission is to revolutionize travel through the power of technology. Twenty years ago, we disrupted the travel industry by turning travel agents’ screens around and allowing customers to book hotels and flights directly. This made customers’ lives incredibly easy because they could take knowledge into their own hands for decision making.


Now, in order to provide natural, contextually aware conversations with our customers we know the best way is to do what we do best, test-and-learn. We began embarking on this journey with Natural Language Processing for the first time about 3 years ago with a few Expedia tasks built on text input. Today, we’re excited to unveil our first foray into voice-activated search content: The Expedia skill for Amazon Alexa.

Here’s a technical look at how it works:

We used AWS Lambda, a zero-administration compute platform, which could not be more simple. It is incredibly fast to get started and scales with ease. A freshly created function is ready and able to handle tens of thousands of requests per hour with absolutely no incremental effort on your part.


Consumers recognize the value of VUIs, and want it to work in more ways. We are committed to expanding and learning within this ecosystem. This is just the beginning of our testing, we are aware that hotel and flight search results remain incredibly complex from a voice perspective. If you conduct a flight search on our desktop, we’ll show you up to 1600 results. Even if we just read you the top five best options, we need to work through the most important data pieces to share with customers.

The future has endless possibilities, and we have a lot to learn, but we are excited to continue to share our progress.