The Liquid Studio Bot is a Google Home virtual assistant guides guests through
Accenture's Liquid Studio,
a studio for demonstrating the latest technology for rapid application development.
The responses of the bot are randomized, so users can
get a different but equally informative response to the same question. The bot is delivered with a simple and
intuitive manual so unfamiliar users can get started with using the bot right away.
As a proof-of-concept, the bot also supports dynamic responses, the data of which is obtained from a database. This database can easily be expanded. If available, the bot will use a display to make the responses to user queries even more informative!
The bot is connected to an analytics system that allows product owners and developers to identify e.g. questions that are often asked and after what questions users often leave the conversation. This helps them to improve the bot where it is most necessary and discover obstacles in the conversation flow. To make future development easier, the bot is delivered with documentation covering all functionality and. This documentation also contains the changes made from the original implementation, if applicable.
While this bot is designed with Liquid Studio in mind, it can be adapted for any institution, for an original and innovative way to introduce visitors to one's company!
Accenture was very precise in what they expected as an end result of the project. All requirements are given, but thinking outside of the box is heavily encouraged. The planning was flexible. Everybody at the studio shows interest in the project and what it can offer.
The team consisted of six computer science students with one specialized in economics as well. We all came to the project with a different skill set. The result was that every member contributed to the project in their own way. This makes our team well-balanced and multi-disciplinary, essential for a great process and project. We learned a lot from each other during the process, mainly when working together on tasks.By seeing each other’s work ethic, every member learned valuable insights into other ways of working. In the end, we worked together as a team and not as six individuals. This results in ideal conditions to excel as a team and produce a product which we are proud of.
Dialogflow was the primary technology we relied on. It allows the construction of conversational
user interfaces, like voice apps and chatbots. In Dialogflow, one can create an agent, which
is trained to handle conversation scenarios using machine learning. The conversations are
handled using intents. Formally, an intent can be described as a mapping from input
sentences to responses. Thus, the agent may receive some input, which triggers an intent,
and chooses the right response accordingly.
The basics of Dialogflow are easy to comprehend, but at the same time, more advanced ideas can be applied. For example, advanced responses can be achieved by integrating a database as fulfillment. For this purpose, we have used Firebase, a NoSQL database designed for fast operations. Different types of data can be saved. For example strings, integers, booleans, but also images.
The last tool we used is Chatbase to collect and view user analytics. There are different types of views that can be analyzed, including the amount of sessions handled, how good they are handled, which flows the conversation follows and how often follow-up questions are asked.