Page 95 - ITU Journal Future and evolving technologies Volume 2 (2021), Issue 5 – Internet of Everything

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 5

The authors in [11] focus on integrating chatbots and IoT applications. The processing time should not increase
to address a critical problem such as air quality aware‑ directly or exponentially with the number of users, but
ness. In this case, the chatbot not only provides users with rather should be constant and perform at its best al‑
information on air quality, temperature and humidity, but most regardless of the workload. To get high scalabil‑
also provides services such as subscriptions to preferred ity, we can rely on serverless cloud services such as Ama‑
air quality monitoring points. Furthermore, advanced zon AWS, Google Cloud Platform, IBM OpenWhisk or Mi‑
functions have been implemented that can be managed crosoft Azure. On these platforms, we are able to develop
entirely via chat such as: alarm services, threshold set‑ lightweight event‑based architectures so as an event can
tings, geoquery and advice based on pollutant levels. have more than one handler and is also able to start the
In [12], a healthcare prognosis chatbot based on AI‑IoT execution of short isolated parts of codes written in order
and with adaptive learning capabilities is proposed. The to perform speci ic atomic tasks. Additionally, each event
aim of the system is to provide medical diagnoses in real handler can create one or more event after processing the
time and to support patients in the absence of healthcare event data. Function as a Service (FaaS) is a cloud service
professionals. The interactive system provides tools to model based on serverless architecture that allows devel‑
collect data, answer general medical questions, provide opers to build a lexible system that its well to pulling en‑
assistance and provide alerts to remind patients that they tire functions up and down for each request. In chat ap‑
need to take their medication. The system in question has plications, the speed with which applications are instanti‑
shown an accuracy of 90% of the answers. Similar to the ated is crucial to reduce latency times. In an FaaS solution,
previous one, in [13] is presented a chatbot designed to the platform manages the loads at the level of individual
increase the capacity of health services so as to reduce the requests, optimizing in terms of performance and costs.
management costs for medical consultancy services. Un‑ However, it is not possible to implement a chatbot system
like the [12] proposal, this chatbot is paired with an IoT entirely in FaaS, as there are other features that require
device for detecting vital signs. This combination can help other service models to ensure, for example, data persis‑
people know their health status. tence, back end to an IoT platform or front end for user
With COVID‑19 social stress has grown exponentially, the interface rendering. And it is not recommended to use
proposed work in [14] uses a chatbot to defeat the stress of exclusively a container‑based service model (Containers
individuals during the period of isolation. This chatbot as a Service (CaaS)) even if currently Kubernetes, at the
allows persons to interface with remote clinical special‑ level of scaling, is approaching FaaS solutions thanks to
ists. In this case, arti icial intelligence and NPL techniques intelligent traf ic management based on analysis models
combined with a clinical chatbot. This will understand if it that imply FaaS features. Based on the application con‑
is enough to continue the conversation with the bot or if the text, however, we can think of a hybrid use of containers
and FaaS, which is the solution we adopted in our system.
user needs to interact with a human professional. From
what has been analyzed, it can be seen that integration with
The use of a pay‑per‑use model reduces operating costs
IoT systems is very useful for enabling inexperienced users
compared to a traditional system that requires the alloca‑
to use advanced features in a simple way. Our proposal is to
tion of the resources of one or more processing instances.
insert a support chatbot within a Social IoT (SIoT) platform.
In fact, the FaaS model allows you to activate the neces‑
In this way, all the applications that will be hosted within it
sary functions on request and to release them immedi‑
will be easily usable and con igurable even by the less
experienced. ately after the execution of the tasks. So we can see that,
given the speed with which requests must be processed
3. BACKGROUND and given the conditions of ic non‑uniformity that
make it impossible to estimate the users who will actu‑
ally request the service, the most convenient solution for
Currently, there are multiple architectural solutions for
the implementation of the chatbot is the serverless one.
IoT (vertical, horizontal, centralized or distributed solu‑
In addition to scalability, it is also necessary to pay atten‑
tions, etc.) with involvement at various levels of the user
tion to the latency at start‑up, that is the time that a FaaS
in interacting with devices that surround them. The de‑
function takes to respond to requests. Typically in all the
sign of an intuitive interface requires key requirements
platforms mentioned they take from a few milliseconds to
that best it the chosen architectural solution. The follow‑
a few minutes, this time is variable and depends on vari‑
ing subsections show the technological needs in the de‑
ous factors, such as programming languages, for example.
sign of a chatbot and the chosen reference IoT architec‑
ture.
Another key requirement is the management of the state,
a serverless system by its nature is stateless, to overcome
3.1 Key requirements in designing a chatbot this problem we will rely on an instance of a database that
system
will take care of saving both the state and further inputs
In human‑interaction‑based applications the processing that will be used to manage the requests.
time and the latency in general are key requirements.
Similarly, in a chatbot application, users expect imme‑
diate responses in comparison to other web and mobile

© International Telecommunication Union, 2021 83