What DocBot is Disrupting
According to the World Health Organization: "At least half of the world's population cannot obtain essential health services. Currently, 800 million people spend at least 10 percent of their household budgets on health expenses for themselves, a sick child or other family members. For almost 100 million people these expenses are high enough to push them into extreme poverty, forcing them to survive on just $1.90 or less a day." Additionally, for many people, doctors are inaccessible due to socioeconomic conditions, locational barriers, and limited doctor supply. Upon learning this when researching Sustainable Development Goals, I experienced a paradigm shift. Basic health services should be a universal human right yet currently, the average person in the world does not have access to healthcare. This is unacceptable and something must be done. One of the most promising existing solutions is telemedicine: Connecting online physicians with patients through text and video conferencing calls. However, there lies a huge problem with scalability and costs. Doctors are expensive to employ which makes telemedicine not viable for the majority of the world's population. Humans physicians and doctors will always be expensive, and scarce to go around equally around the world. We needed a solution that is a viable option for everyone in the world, no matter the background.
How can we connect essential health services to half the world that lacks it?
The current best solution is telemedicine, connecting doctors to patients through telecommunication, which is not an affordable option for billions of people. But what if we could maintain the quality of telemedicine while decreasing costs and increasing accessibility by 100X without the employment of doctors? This is the mission of DocBot, an AI chatbot that eliminates the need for telemedical doctors. Unlike doctors, DocBot is significantly cheaper as the costs lie in cloud computation which costs pennies per session. DocBot uses Natural Language Generation to understand and respond to patient responses indistinguishable from a human doctor. To deduce symptoms, DocBot uses Natural Language Processing to identify and classify relevant information from the conversation. Once all symptoms have been communicated by the patient, DocBot will run the classified symptoms through APIMedic API to accurately diagnose the patient's condition and prescribe treatment accordingly. DocBot uses both generative (to respond to anything the patient says) and retrieval based chatting (allows the integration of preprogrammed responses: Searching nearest pharmacy on Google Maps, APIMedic for diagnosing, accessing data). V1.0 of DocBot (https://youtu.be/LGWlaNZBScc), used a Tensorflow :) Sequence to Sequence Recurrent Neural Network trained on a conversational dataset to generate human-like responses. It has several preprogrammed functions, showcasing the ability for retrieval based chatting. Additionally, by implementing Text2Speech and Speech2Text as well as Natural Language Translation APIs, DocBot can be used by everyone, as long as they know how to speak a language. In the development is V2.0 of DocBot (mockup attached below) which uses a PyTorch NLP Model trained on a dataset of Doctor-Patient Dialogue, making the generated replies much more relevant. DocBot V2.0 uses Google Maps API to search up nearby pharmacies and can diagnose condition + prescribe corresponding medication using MedicAPI. Once the backend of DocBot is finished, the potential frontend applications are endless. The goal is to make it as accessible as possible to people of all backgrounds. DocBot would be made into a multilingual mobile and web app, allowing anyone with a device and internet connection to receive telemedical health services. Furthermore, DocBot can be easily deployed onto existing telecommunication platforms like Messenger and Slack. In the near future, DocBot will come with an API, allowing developers to integrate Health Services into applications. DocBot will be accessible from nearly everywhere on earth- all one would need is a device and internet connection.