Sentiment versus sentient: Taking conversational AI further in travel

Artificial intelligence applications that are currently being used in the travel and hospitality industries boast a myriad of benefits - from booking flights and hotels to organizing a traveler’s itinerary or suggesting places to visit - all through the platform of your choice.

The beauty of these AI bots is that they’re accessible 24/7, support different languages and give immediate response to queries.

While most of these applications in use today already understand simple conversational dialogue - like basic Q&A and form-filling - how do we get to the next level of more sophisticated dialogue that helps chatbots process who a user is and personalize the conversational experience to their unique needs and current sentiment?

The next generation of conversational AI is complicated but within reach for hospitality and travel brands that want to deliver superior guest experiences using natural language chatbots.

Thanks to voice fingerprinting and Long Short-Term Memory (LSTM), tomorrow’s AI will have no trouble remembering your needs over the needs of its many other user-based interactions

Madhusudan Mathihalli

Share this quote

To have a successful exchange with a user, AI chatbots have to master several different types of observation.

Planning, booking and servicing travelers demands visual feedback and specific answers that can’t be attained through voice interfaces alone - for example, browsing through photos to choose a resort stay, filtering through flight options, or watching a 360-degree video of a property.

Because of this challenge, it becomes imperative to layer in next-gen technologies to provide more context and better identify a traveler or guest’s specific intent.

The more information we can capture through different sources, the better the AI’s dialogue with a customer will be.

In addition to an unprecedented level of personalization, these additional contexts enable the next generation of conversational AI to negotiate.

Consider a “natural” dialogue for selecting a flight. “Would you rather I find a flight that is cheaper in price, or closer to a 10 am departure time?”

You may change price, airport or airline, which requires a bot to recall and distinguish accurately between multiple variables and negotiate a solution that will fit the criteria you’ve established.

It will even be able to read the temperament of a customer through their facial expressions using facial recognition.

If a chatbot can detect that you are frustrated based on your facial cues, it could more quickly direct you to a live agent.

Arguably the most impressive capability of this next generation, however, is reference-based dialogue. Imagine the ability to say, “Book that flight we talked about yesterday,” and the bot will recall all the relevant previous context and finish out the transaction successfully.

Thanks to voice fingerprinting and Long Short-Term Memory (LSTM), tomorrow’s AI will have no trouble remembering your needs over the needs of its many other user-based interactions.

The biggest challenge the industry faces in bringing these sophisticated technologies to fruition involves the subtle nuances we make as humans in everyday conversations with each other.

Slight variations in our speech patterns and the ability for a bot to distinguish between one user and another are especially difficult. But despite those struggles, the travel and hospitality industries are already employing and testing out solutions.

A leading luxury resort and casino in Las Vegas sought to seamlessly connect its guests with personalized, instant recommendations and bookings of its amenities, including DJs, restaurants, shows and more.

Through a natural language processing chatbot deployed on Facebook Messenger, guests can now discover relevant hotel information and activities and book and pay for services all through the chat interface.

With the emergence of Alexa and voice platforms, hotels are integrating these capabilities into in-room interactive voice response concierges to provide true 24/7 service.

In order for the conversational components of these interactions to mimic the types of dialogue they would have with a live human, let alone a five-star concierge, AI must employ the three-dimensional holy grail of voice, face and dialogue to understand context and intent.