Students from Haaga-Helia and the Higher Institute of Engineering of the Polytechnic of Oporto, Portugal, implemented chatbots together in a course on AI offered by Haaga-Helia. The project idea and the access to the IBM Cloud environment came from the Finnish company Hotelway.ai. The European cooperation was organized through the Dihub-project.

Flexible Chatbot Implementation using Cloud Services

Chatbots are computer programs that communicate via natural language in a synchronous manner with a human (Russell and Norvig, 2020). The communication may happen in a written form or it can be speech. Chatbots may be implemented using cloud services hosted by a cloud service provider or on premises by using own servers and an existing chatbot framework. In the case of cloud services, chatbot development environments are typically also offered as easy-to-use SaaS (Software-as-a-Service) products.

The aforementioned two alternatives to chatbot development: 1) cloud services and 2) the private environment, both have their pros and cons. The main differences in the approaches are the upfront costs when starting the development work, the level of technical skills needed and data privacy.

Cloud services offer a flexible and cost-effective possibility for experimenting with new technologies as the pay-as-you-go pricing model only bills for those resources that are actually used. In the experimenting phase the usage of resources is typically low because the number of end users, concurrent processes and the amount of data are small. In the case of a SaaS environment, technical skills for the installation and administration of software frameworks are not needed as the service provider takes care of those tasks.

Data privacy is an issue when using the SaaS model. All data is hosted by the cloud service provider and thus the chatbot developer has to ensure that the SaaS service is compliant with the requirements of the GDPR, among others.

Well-known commercial cloud-based products for chatbot development include: Google Dialogflow, Microsoft Bot Framework, IBM Watson Assistant and Amazon Lex. These frameworks offer seamless and simple integration with a variety of text and speech based communication platforms such as: Facebook Messenger, Slack, Twitter and Twilio.

If one wishes to have full control over the environment, it has to be hosted on premises. The leading solution for this is the open source chatbot framework Rasa. Another well-known open source platform for chatbot development is the Microsoft Bot Framework. However, the natural language understanding service (LUIS) behind it is not open source.

Course Project on Chatbots

Approximately one half of the AI course is project work. The project is typically a real-world task given by a company. In the fall 2020, it was a chatbot case given by the Finnish Start-up company named Hotelway.ai.

As materials the students got about 250 questions and answers and free hands to design and implement a chatbot. The topic of the chatbot was Finland and the Finnish Pavillion at the World Expo in Dubai. For inspiration the students took a look at existing chatbots in the field of leisure time and travel. The students did get both theoretical training on chatbots and practical guidance on the use of the tool – the IBM Watson Assistant.

The IBM Watson Assistant is a SaaS product where the developer may design dialogues by creating intents and entities. The answers that the bot typically gives contain text, images or further questions. Even though the data given to the students upon project start only consisted of questions and answers, the students were encouraged to build dialogue into the questions.

The task was not at all straightforward. The students had to merge and separate some questions, design a dialog and decide upon the relevant intents and entities, among others. At the same time, they were learning about the capabilities of the IBM cloud service platform.

The projects resulted in three excellent chatbots, one of which you are invited to test here!

Conclusions and Future Perspectives

The chatbot course project was the first in its kind: it combined the teaching of cloud-based AI technologies and a real-life company project in a European setting.

The aim of the project was to increase awareness on cloud-based AI technologies and to provide students with skills and knowledge relevant in the job market. Another goal was to foster co-creation and the sharing of knowledge related to cloud technologies on a European level.

The Dihub project will continue for one more year and the next implementation of the AI course will be in the spring term 2021. We are looking forward to continuing the European cooperation on innovative cloud-based AI projects!

Further reading

• Russell, S., & Norvig, P. 2020. Artificial intelligence: a modern approach. Third Edition, Pearson.