The Health Research Institute of the Balearic Islands (IdiSBa)

Yed.ai allowed any instance of the local Government to collect, identify and respond immediately to the population's main doubts in relation with the Sanitary Crisis, 24/7.
75,7%
Neural Network efficiency after 4 weeks
64%
"Useful" answers rated by users
26 000+
different questions identified

" YedAI the chatbot we relied on to inform about COVID19 pandemic, helped us to organize the information that we used to answer citizen's questions at our phone line. We trusted YedAI to digitally reach populations who usually don't use official phone numbers. YedAI has already changed the way we work. We are very grateful to the team! "

Face
I am María from Health Promotion Department
The challenge

Question: Is it even possible?

The Health Research Institute of the Balearic Islands (IdiSBa) opened a call for proposals in order to allocate European Union and local funds to innovative projects looking into the research and development of solutions to combat SARS-CoV-2 spread and manage its consequences. At that time YedAI was a prototype of a cloud-based web chatbot using a bespoke framework of Machine Learning techniques and was presented as project proposal for a Minimum Viable Product that could serve the Public Health Administration in its efforts against the virus.

The proposal was approved by the IdiSBa experts panel with the condition that the conversational bot would be implemented in collaboration with the Balearic Islands' General Department of Public Health and Citizens Inclusion who put together a team of nurses, doctors, social services, psychologists and public health technicians to offer a COVID-19 free telephone line service within office hours to help resolve all citizens' doubts related to the pandemic.

Curated & Supervised Answers

A matter of state importance

It was of utmost importance to provide accurate information to citizens and the government capability of managing the crisis was at stake. Therefore, the content creation, knowledge management process and copy skills of the team were as important as the neuronal network, the quality of code, the set up of the Amazon Web Services stack and security measures.

The solution

Meet the Botibot

The conversational bot was named "Botibot": a simple mobile friendly website with an avatar wearing a facemask was created and the implementation of YedAI allowed any instance of the local Government to collect, identify and respond immediately to the population's main doubts in relation with the Sanitary Crisis, 24/7.

Screenshot illesbalears.botibot.es

YedAI team of experts assigned for this case included a PhD in Physics, a Computational Linguist, a Language & Literature content editor native speaker in both chatbot languages, a Software engineer, a Systems Reliability Engineer specialized in Cloud & Security, a Fine Arts Graphic designer and an Agile Coach - Product development expert. The government information was structured and analysed by the team so that it could be evolved and updated everyday / week according to the COVID-19 situation for each of the 4 Balearic islands that form the archipelago (Mallorca, Menorca, Ibiza and Formentera). The information was related not only to Health topics, but also about Public aids, Gender violence, Work & Employment measures, Mobility & Transportation, Tourism as well as Food Safety. In this Machine Learning application, the bot learning process was fully supervised.

The results

Data Science expertise is key

The effectiveness of the neuronal network is related to the machine learning techniques application. This means that Data Science expertise is a key factor for the efficacy indicator final result.
Question mark sign

26,000+

types of questions answered

The final Yed.ai version was able to answer more than 26,000 different questions, identifying the user specific need and displaying the most probable answer within the Public Health ad-hoc created knowledge base.

alt

75,8%

network efficiency after 10 weeks

This indicator sums up the "Correct" answers 65% plus the "Not available/known" answers 10,8% . The neuronal network capacity of recognizing what the bot should not answer is also considered an efficient result.

alt

64%

“useful" answers rated by users

Botibot was the only chatbot of its kind within Spain that requested users for their direct feedback once each response was provided, the team was able to evaluate whether the answers were really useful from a citizens' perspective.

People is the most important

User feedback was taken in real time, and also collected via social media channels where the bot service was launched to the general public. Content was adapted daily according with pandemic new information available and the new containment measures taken to prevent virus spread. The continuous feedback loop along with the engagement of the Public Health technicians in the evaluation of the content usefulness and eagerness to serve the citizens the best way possible through clear information, was probably one of the most motivating aspects of this project for the core team. Botibot COVID-19 success case illustrates how a human contact centre team managed by a regional Government is tied together with an AI-powered chatbot to get the most effective mix to handle an information service with high quality levels in the G2C arena.