How conversational AI is changing communication in tourism

In view of the shortage of skilled workers and digitization, the need for intelligent search functions and chatbot solutions is increasing. If the applications are based on well-structured and open data, conversational AI could transform communication in tourism.

For many people, especially younger ones, the use of voice interfaces is now taken for granted – typing is increasingly perceived as a nuisance. Accordingly, access to tourism data and services in natural language has a high future potential for tourism applications supported by artificial intelligence (AI). It is expected that intelligent voice modules will find their way into all kinds of applications, e.g. voice assistants such as Alexa and Siri in the home or office, in dialog with assistants in the car (with similar dialog functions to voice assistants in the home environment), telephone assistants for customer service (first-level support), chatbots and other applications that are still being tested in the tourism context today, such as the Metaverse or AR/VR applications.

Tourist services are hard to grasp and cannot be tested in advance, which is why guests have a greater need for information before they travel. Offers or activity planning during the vacation are often planned by the guests themselves or on the recommendation of a tourist information, so that customers are active co-creators of their vacation experience. This leads to the fact that tourist services are description-intensive and this information is an important aspect of the travel decision. In view of the increasing shortage of skilled workers and digitization as a whole, the need for intelligent search functions and chatbot solutions that can answer simple queries quickly and accurately is growing accordingly.

Consequently, access to tourism data and services will increasingly be through natural language in the near future, as voice recognition and playback technology has matured and voice exchanges are part of our natural human interaction and communication. Control elements such as keyboard or mouse can therefore be seen as bridge technology. Web browsers and other applications already support voice input. Processing complete correct sentences often yields better results in Google than pure keyword-based searches. Before we look a bit behind the scenes of the data and underlying technology, a practical example:

Finding a suitable bike tour is still a challenge for guests. The simple question: “I want to take a bike ride from here tomorrow. It should last a maximum of three hours and provide a nice play opportunity for the children along the way. Afterwards, we would like to go out for a delicious ice cream.“, can be answered by a tourist info employee in no time, but requires several Google searches or the use of specific apps for the guest. The results are often incomplete and partly outdated, e.g. when asking whether the ice cream parlor is really open.

So how can a conversational AI answer this question?

A conversational AI can decompose questions into their elements of meaning (search intentions), as described above, as well as recognize their interrelationships. This works all the better the more extensively the AI has been trained in advance with the corresponding factual knowledge (e.g., from a knowledge graph). The answer in turn can be generated from a Knowledge Graph. Knowledge Graphs, as Google shows, provide a suitable technology to describe and link data in a simple and flexible way and to represent this factual knowledge consistently through uniform semantics.

To obtain the most plausible data possible, it makes sense to collect the data locally and aggregate it in a knowledge graph. AI algorithms can then be used to further improve data quality:

  • Verification
    Address data can be verified, for example, by matching it with other geodata sources. This increases the machine processing quality and makes it possible to always answer guests’ questions correctly and reliably based on the verified data.

  • Enrichment
    In addition, further details, such as frequently asked questions, for example whether a bike path is better to ride with a mountain bike or a road bike, can be added later.

At highly frequented locations, supplementary information on capacity utilization (live or in the form of forecasts) would be useful, as is already currently being done in the nationwide joint project AIR (AI-based recommender for sustainable tourism) and in the similarly designed project LAB-TOUR SH (statewide digital visitor management for tourism in Schleswig-Holstein).

For such applications, a uniform data “language” in the form of open and interoperable data models that go beyond simple exchange formats is imperative. This is precisely the approach of many nationwide and European initiatives that are working to develop standardized open data models that provide the basis for conversational AI to always make use of the data it needs for the particular search intent.

The use of voice assistants and chatbots is manifold: the website can be equipped with these tools, as it is already used as a point of contact with guests. During the trip, the voice application can be installed on access points of public WLAN or act as a voice assistant in the tourist info. The voice assistant will also answer questions at midnight and communicate details about the weather or tours for the next day. Those guests who prefer to do their travel planning in the Metaverse or similar virtual applications in the future will enjoy the interactions with the avatars. Output channels can be combined to provide the best possible experience. The complementary on-screen display of a bike tour suggested by a voice assistant can be quite helpful.

However, any chatbot or voice assistant will only be as good as its underlying data. Quality and timeliness are an important factor. Due to the increasing machine processing of data as well as their linking and evaluation, internationally established data models that guarantee uniform semantics are important components.

Portrait Prof. Dr. Michael Prange

Prof. Dr. Michael Prange

Kiel University of Applied Sciences

Prof. Dr. Michael Prange is Professor of Data Science and Project Manager for the collaborative projects AIR and LAB-TOUR SH at Kiel University of Applied Sciences.

More about the person at: