Page 9 - AeM_April_2019

RESEARCH, ANALYSIS & TRENDS

be optimized - from product evaluation, the purchase to website. The Bot Framework SDK makes it possible to
service - through greater efficiency and convenience. implement the bots. The Language Understand
Besides algorithms that control communication through Intelligence Service (LUIS) assists the bot with deep
keywords and communication patterns, artificial learning and linguistic analysis to develop a common
intelligence is being used to learn from preferences and language between machine and humans and bot
interactions. This allows systems to better adapt and connectors allow the bots to be integrated into different
proactively control communication. messaging platforms, while the bot directory allows the
distribution and discovery of other bots. We can expect
With advances in artificial intelligence and machine that the bot development will lead to fundamentally
learning, more challenging tasks can now be performed different principles in communication and in the
by bots. And again, even the idea of the bot isn’t new corresponding interfaces. Bots will replace most of the
and actually dates back to the 1950s, when Alan Turing, websites and apps and make the separation of
an early researcher in computer intelligence, presented application-related functions void. Thus, a transaction
an attempt to test the intelligence of machines. Known may include the evaluation of a product, the selection,
today as the Turing Test, it works that way that if more as well as the purchase and service.
than 30 percent of a test group is convinced that they
are conversing with a human and not a computer, the Typically, a consumer would need to use various apps
machine is assumed to have a human-level mind. and/or websites to do that. The bot, as a kind of
operating system, however, combines the different
A small breakthrough came in 2014, when a third of the forms of information and interaction into a continuous
participants believed they had a conversation with a transaction: the bot has selected, according to the
human even though a bot was used. And let’s be honest, preferences learned, has triggered the order, and
nowadays, it is not always easy to make the difference completed the transaction via the known bank and
between man and machine in a conversation. Even address data. If, so far, bots have been quite trivial, this
comparatively little artificial intelligence is enough to will change forever in the age of artificial intelligence.
mimic the illusion of a natural human interaction. In this Previous implementations resorted to internal
regard, however, the developers of bots face many databases, keyword-tagged phrases, and developer
challenges. Their goal is to develop a common language rules, but now the bot scans the customer input for
between machine and human to facilitate keywords, then compiles the knowledge and text
communication. For businesses to be able to offer their modules according to the fixed rules and returns the
services on messaging platforms, there must be generated output to the customer. Extensions of the
programming interfaces (APIs). The APIs allow that an system - in the form of new knowledge, linking of rules,
external program code, such as a bot, can be integrated keyword tagging and text modules - must be
with existing software, such as a messaging platform. programmed.
But not all companies have the expertise to build their
own bots and integrate them into a messaging platform, Today's bots also use the largest available dynamic
hence, there will be most likely a ‘bots-as-a-service’ database in the world, which is the Internet, and the
concept available soon, that simplifies the development semantic web facilitates automated access to
and integration of bots. The developed bots should be knowledge. By using the customer's interaction
both universal, which means easy to use on all different patterns, the bot can make use of the deep learning
platforms, and easy to build. algorithms of machine learning to find custom keywords
and maintain its own custom and automated database.
Two such bot builders are already available on Not to mention that recent breakthroughs in Natural
Facebook and Microsoft Language Processing (NLP), the AI subsection that
deals with human-machine communication, boosts bot
Already 3 years ago, Facebook reported on its developer development even further. In fact, the more intelligent
conference that it has created new interfaces to form of bots secure themselves already information on
Messenger for external developers. Wit.ai, a software their own from online sources and combine them to
that helps develop an API for voice-enabled user new content. The AI-based bots are also fed by the
interfaces, was previously connected to Facebook to answers and reactions of the users - which certainly
help developers integrate their services. In the first two reduces the ability to control the information used for
and a half months after the messenger's approval, more learning. Will chatbots take the lead in one-to-one
than 23,000 developers signed up with wit.ai and over conversations on the internet? We’ll see! ◊
11,000 chatbots were created. The messenger now also
provides a visual interface to enhance the user By Daniela La Marca
experience and includes plug-ins that can integrate the
bot into third-party offerings. Since autumn 2016, it is
also possible to handle payments directly via the
messenger.

If the credit card information is stored in Facebook or
Messenger, the transaction can be made without further
input.

The Microsoft Bot Framework provides good conditions
for developing bots for different platforms or your own

MediaBUZZ Pte Ltd - Independant ePublisher for Asia