We talked to Zaynab DIALLO, who worked as part of the ‘Let’s Break the Cycle’ consortium to create the Weerwi chatbot. Zaynab is a former student of the Dakar Institute of Technology (DIT), and a PhD student in statistics at Ohio State University, USA.
Before talking about the Weerwi chatbot, can you tell us what a chatbot is?
Chatbot is a word that we’re hearing more and more frequently and that’s becoming part of our daily lives. It’s a software application (bot) used to conduct an online conversation (chat) via text or voice synthesis, to complement or replace contact with a live human agent.
You may have already encountered a chatbot when you needed information about a website or when you were trying to contact a company via a messaging application such as Whatsapp or Messenger. Chatbots are part of a large family known as conversational agents, which are computer programs capable of conversing in real-time with a user in natural language, either written or spoken. The aim is to make the conversation as natural as possible so that it is indistinguishable from a real person. They can be used on a website, on a messaging application, or by telephone.
These chatbot-based discussions can enable users to quickly find a solution to their query. However, if the chatbot is unable to answer the question, it often redirects the user to real people.
There are several types of chatbot, depending on the need. There are the rather simple ones, which operate on the basis of pre-established rules during their development. They have pre-recorded responses and guide the user through a menu/list of choices to arrive at the desired response. And those that use Artificial Intelligence techniques, which can become very complex. They learn from existing conversations to understand the user’s context and intention, enabling them to handle new situations outside their program.
For companies or organisations, providing their customers or users with a chatbot can add value to their service. For example, to take charge of answering certain recurring questions that are easy for a chatbot to manage.
In the case of Weerwi, a chatbot integrated into its application, website and Facebook page can provide first-level guidance 24/7 before a professional is called in when necessary.
Why did you decide to create a chatbot for menstrual health?
We were very enthusiastic about the idea of creating a chatbot for Weerwi because menstruation is a subject that affects the vast majority of women. However, in Senegal, as in many countries, talking about menstruation is still considered taboo. Many young girls don’t have access to enough information when they first start menstruating. This lack of information is the result of a desire to “protect” young girls, making menstruation a mysterious phenomenon, verging on the mystical, when in fact it is a natural phenomenon. We believe that women need to understand what’s going on inside them, so that they can feel good about their bodies and take care of their health. We thought a chatbot could help with this.
In designing the Weerwi chatbot, our aim was to communicate unbiased, scientific information, but also messages of support to young girls and women so that they can live their menstruation to the fullest.
How was the Weerwi chatbot created?
The Weerwi chatbot was designed to answer users’ questions about menstrual health. It was designed in several stages:
To begin with, with the project’s stakeholders, ApiAfrique and IT4Life in particular, we listed the topics that might be of interest to users (periods, menstrual cycle, period management, advice, taboo subjects, etc.) and identified their sub-topics. We then put together a list of questions and answers for each sub-topic and defined the flow of conversations between the Weerwi chatbot and a potential user. The aim is to answer her questions while suggesting other topics that might interest her. This is a critical phase that requires us not only to anticipate the user experience but also to define the possible links between different themes. This stage ended with a review of the content by a qualified gynaecologist to ensure that the information provided was reliable.
We then launched the development of the tool in the form of an application on DialogFlow CX, a Google tool. Development was carried out through iterative improvements, and it was only after testing with target users, to get feedback and incorporate their recommendations, that the final version of Weerwi’s chatbot saw the light of day.
Today, the chatbot can be accessed in several ways: on the weerwi.com website, on Facebook messenger via the Weerwi page and directly in the Weerwi application.
How does the Weerwi chatbot work?
When a user interacts with the chatbot from one of the platforms mentioned (the Weerwi app, Facebook messenger or the website), her message is sent to the chatbot application, which is integrated into DialogFlow CX using API calls. The APIs act as an intermediary between the platform from which the message originates and the Weerwi chatbot. Thanks to the settings already made, the chatbot will “translate” the user’s message at each turn of the conversation and then provide the corresponding response. This answer is sent back to the platform where the user can view it on her screen.
What challenges did you face in creating the chatbot? What feedback do you have on using the chatbot?
The first challenge was to find the right words and explanations that respected the sensitivities and culture of the users. Our aim is to inform and not to scare women. We also wanted the chatbot to be a reassuring tool that gave reliable advice without giving the impression that it was taking the place of the medical profession. That’s why you’ll often see in the chatbot an invitation to talk to a midwife (link to the Lydia Conseil call centre at partner DKT) or a recommendation to go and see a doctor.
Another challenge was to put ourselves in the shoes of a user of the Weerwi chatbot by imagining her questions and behaviour – because the use of chatbots is not yet widespread in Senegal. Once the chatbot had been created, we had to make sure that it really met the needs of our target audience. To do this, we recruited young girls aged between 14 and 19, who tested it. They gave us feedback on their experience and gave us their opinion on the changes or improvements required.
So far, the feedback from the Weerwi chatbot has been positive. There seems to be a real need for it, since it was contacted more than 35,000 times in the space of a month at the end of 2022. In the future, at the request of many users, we plan to offer a version of the Weerwi chatbot written in Wolof. In the longer term, we hope to be able to develop an audio version.
What did you personally learn in the process of creating this chatbot?
As a woman, I learnt a lot about the menstrual cycle during this project. From a technical point of view, I developed a chatbot outside a test environment, which was a first for me. During the DIT bootcamp, I learnt the basics of developing a chatbot, and this project enabled me to apply them and surpass myself. Furthermore, this project confirmed my strong interest in teamwork. The ApiAfrique and IT4Life teams, in particular Marina Gning, Soukeyna Ouedraogo and Kathleen Chau, were essential and showed great responsiveness in building the Weerwi chatbot workflow.
Interview conducted with: Zaynab DIALLO, a former student at the Dakar Institute of Technology and a doctoral student in statistics at Ohio State University, was asked to create the “Weerwi-bot” chatbot as part of Breaking the Cycle