Travel operator Thomson has created a chatbot using IBM Watson, combining machine learning techniques to provide holiday recommendations to customers.
The virtual assistant search tool, which will be available to customers as a beta pilot next week, will offer suggestions for travel destinations using Watson’s natural language processing and speech-to-text capabilities.
For example, a customer might say that they want want to visit local markets on holiday and recommendations will be made for related holidays.
Read next: What is a chatbot, how do you build one?
According to Jo Hickson, head of innovation, at Thomson’s parent company TUI UK, the popularity of personal assistants such as Apple’s Siri, Microsoft’s Cortana and Amazon’s Alexa has raised expectations for search features on websites.
“The sophistication that’s happening with offsite search has to be considered in terms of what an ecommerce retailer site should do and be in the future,” said Hickson.
She said that the search bars, text boxes and filters on ecommerce sites require “fairly high effort on behalf of the customer”, and Thomson wants to make it “as easy as having a conversation to get a short list of options”.
Hickson said: “Because our inventory is large we have lots of holiday options, we go to many destinations around the world and then even within those options there are lot of filters and choices, depending on whether or not you are a family, what time of year you want to go etcetera.
“So this is all about reducing customer effort and making it frictionless and modern.”
The chatbot was built by IBM using its Watson cognitive computing platform and cloud services. The process involved inputting large volumes of TUI data related to travel destinations and will be optimised by incorporating user reactions to improve the accuracy of recommendations.
“We are also bringing in how the users react to the site and using that in combination with the original data to get the machine learning [capabilities],” said Holly Cummins, technical lead at IBM’s Bluemix Garage in London.
“If people say 'I'd like to lie on the beach' and we suggest Tenerife but they never click on Tenerife, we might learn that the reason that people choose to go to Tenerife isn't really the beach, it is something else. So we are starting with the seed of the TUI data and then people interact with it and then it gets smarter.”
Hickson said that there are other potential uses for the technology at TUI and Thomson in future. For example, natural language processing could be used as part of its contact centre operations for certain queries.
“We have a contact centre - and we will always have one because customers need to contact us - but a good percentage of customers contact us about things that could be answered by an intelligent solution like this,” she said.
“Customers will call us about things like luggage allowance which is a ‘yes/no’ absolute, that is the kind of question that in the future could be answered by this kind of technology.”