This article outlines the process of crafting Quicktext content and delves into the underlying reasons for our approach.
Quicktext employs the so-called "dialogs" to gather information and data about properties, ensuring that the data entered into these dialogs ultimately guarantees the quality of answers provided by Velma to users.
1/ How Quicktext creates dialogs
All dialogs revolve around the services/facilities/amenities of the property.
Dialogs usually contain a set of data points, each data point covers and conveys specific information. In other words, a data point is an element of response to a specific question or inquiry.
Example:
Dialog 10-01 check-in time: Evidently, this dialog covers the check-in topic.
At 1st glance, it seems quite simple, however, experience showed us that the check-in topic is more intricate than it looks as users don't only ask about the standard check-in time but often they inquire about the possibility of early check-in, the check-in supplement if applicable or whether they can use the facilities of the property(Bar, swimming pool, spa, gym..) before check-in.
We crunch data from the pool of information available to us from thousands of conversations that Velma generates across all our customer's accounts. These data are:
- Proofread
- Fine-tuned
- Analysed
The data analysis conducted on the collected data determines:
A- What are the new dialogs/topics that must be built from scratch
B- What are the existing dialogs that need to be reconfigured/overhauled/enriched
Examples:
- New dialog added through data analysis: 13-32 Medical services
Regular data analysis has shown that questions about medical emergencies keep popping up, thus, we concluded that Velma must be able to respond to this kind of inquiries shortly. As a result, we began working on a new dialog/topic that we dubbed Medical service. - Reconfigured existing dialog: 15-51 Welcome Tray
This dialog has been renamed and reconfigured following once again an exhaustive data analysis. The welcome tray used to be called a Welcome platter. The old configuration only stated the list of items available in the welcome menu of the room regardless of the room type/category, the refill frequency, the price if applicable, and whether the price is applied per day, per stay, or serving.
Data analysis demonstrated that more data points should be added to enable Velma to provide more accurate and detailed answers to users. Answers that are more in line with today's hospitality landscape.
2/ How Quicktext dialogs are built?
Building a new dialog could be a daunting task sometimes. Luckily we are a data-driven company.
We use data analysis all the time and for everything and building dialogs/topics is no exception.
As mentioned above, data analysis helps us to determine what are dialogs to be added or the dialogs to be reconfigured but it is also crucial in the conception and building phase itself.
Building a dialog implicates working on 2 fronts:
A- The AI front:
When working on creating or reconfiguring a new dialog/topic our AI team uses ALL the questions, jargon, and terminology from actual conversations. In essence, they inject all this information into Velma's algorithm(Q-Brain) to ramp up her understanding abilities as our long AI experience taught us that humans express their ideas in very diverse ways, thus, a chatbot that is only able to understand keywords will not live up to the challenge. At Quicktext we focus on context, a strong AI is an AI that can correctly understand, interpret, and decipher complex intents. Keywords can be sometimes misleading since they don't always express the true intent of an idea.
B- The Dashboard front and chatbot front:
In parallel to the AI front, we work simultaneously on the back-end configuration. Once again here data play a key role: The choice of every data point is based on data analysis.
Example: The configuration of dialog 13-55 - Shuttle.
Every data point was added because we found questions from actual users seeking answers to specific questions==> The elements of response to these questions are none other than the data points.
Data points:
- Destination: Users were asking Velma what is the destination of the shuttle service
- Duration: Users were asking Velma how long is the journey.
- From- to: People want to know what is the timetable of the shuttle service
- Every: People want to know what is the frequency of the shuttle service
- Price: People want to know whether the shuttle is free of charge or paid and how much is the price if applicable
- Location: People want to know what is the exact pick location
3/ Why Data analysis is a guarantee for success?
Imagine as a hotelier you can build your dialogs/topics, and add Q&A and scenarios at will.
At first glance great isn't it? Well, it's not.
If you get free reign to build your dialogs/topic/Q&As, you take out of the equation all the AI groundwork mentioned above, thus, your chatbot will not have strong understanding and interpretation capabilities, as a result, any question or query that is formulated with different words than the scenarios you built won't be understood by your chatbot==> Frustration machine called chatbot.
Conclusion
While our approach may appear rigid initially, lacking the perceived flexibility some hoteliers seek, a deeper understanding of AI technology reveals that this apparent inflexibility is the cornerstone of guaranteed success. The usage of structured data is underpinned by two key factors: precision and relevance in AI-generated responses and time savings in data configuration.
A robust AI-powered tool demands a substantial volume of structured data and seasoned data experts to effectively leverage that information, ensuring optimal performance.