AdMed unlocks answers to health questions with Watson Assistant

By | 5 minute read | November 24, 2020

It can be challenging to communicate accurate scientific information, especially in times where the COVID-19 pandemic is by far the most searched topic on Google, impacting everything from our news to the economy. Having so much information available also leads to the rapid spread of all sorts of misinformation, including outdated guidance or unproven treatments which easily go viral only to worsen the already existing public health crisis.

Similar to COVID-19, there are a number of other public health issues needing attention, including the opioids epidemic. According to the Centers for Disease Control and Prevention (CDC), between 1999 and 2018, almost 450,000 people died from an overdose involving any opioid, including prescription and illicit opioids.

In response to this reality, healthcare agency AdMed, Inc. partnered with the IBM Data Science and AI Elite (DSE) to create the ultimate learning experience for crucial healthcare topics like this, making information available to a variety of audiences — ranging from sales teams and healthcare professionals, all the way to consumers. By leveraging the IBM Cloud Pak for Data Platform, AdMed opened a new umbrella of possibilities for smarter and more interactive learning.

Behind the science

AdMed is an integrated agency with a mission to ensure healthcare professionals, sales teams, staff, and consumers understand the science behind their products. AdMed makes science simple with their two-minute video lectures presented by doctoral-level scientists who educate their audience on new medicines to treat disease and improve quality of life. AdMed is coupling its visual learning experience with an interactive interface, made possible by IBM Watson, to cover a variety of topics, including the science behind:

  • COVID-19: Understanding the latest information on the virus, the mechanism of action, and upcoming treatments under review
  • Opioid addiction and epidemic: Uncovering the neurobiology of addiction/substance use and what we can do to prevent the next generation from suffering
  • Women’s health: Exploring diseases and treatments to improve quality of life

Just ask Watson — I mean, Opie

AI can certainly help make learning experiences smarter. Imagine you could ask questions to a virtual teaching assistant and get not only the answers, but also references to videos and playback containing the searched topics. This is exactly what Watson Assistant enables, which led to the creation of Opie, the cutest virtual assistant in town.

Joan Francy, CEO of AdMed, shares: “We found Watson Assistant to be easy to use and very scalable. The interface allows anyone to create a chatbot, while also enabling our developers to leverage the full power of Watson.”

Powered by the latest natural language processing advancements from IBM, Watson Assistant allows for building, training, and deploying conversational interactions into any application, device, or channel. Most chatbots try to mimic human interactions, which can frustrate users when a misunderstanding arises. Watson Assistant takes it further. It knows when to search for an answer from a knowledge base, when to ask for clarity, and when to direct users to a human.

Watson Assistant makes it easy to train a virtual assistant

Question answering systems and virtual assistants are a very interesting topic in AI research. The goal is to answer any question related to the knowledge base with which the system is trained. But humans can ask questions in many different ways. The question “Can animals transmit COVID-19?” could also be phrased as “Could I get coronavirus from my dog?” The virtual assistant should provide consistent answers for these two cases, and ideally for any other phrasing of such question.

As with any AI project, training data is key for a successful virtual assistant. Given a paired list of questions to answers, one common technique is to take the input from the user and find the closest question in the training set, then returning its corresponding answer. The challenge lies in how we define ‘closest.’ This involves (1) a distance function — like cosine similarity — and (2) a feature space — such as TF-IDF, Word2vec or BERT.

This may sound like a lot of work — and it is — which calls for deep collaborating with several parties involving data scientists, developers, designers, and subject matter experts. Thankfully, Watson Assistant makes this easy and efficient by bringing years of NLP research and development into an end-to-end question answering solution that can be quickly deployed. This includes cloud deployment, user interface, and integration with web and mobile platforms.

The best part? You won’t have to write a single line of code and you can build an assistant in a matter of minutes.

Train your own chatbot free and fast

Francy reflects back on AdMed’s time spent with the DSE, sharing that this partnership “has allowed AdMed to grow our skillsets within Watson and AI technology in order to develop products to enhance our business growth. Oscar was able to break down the steps in a simple way so that we were able to build our chatbot to help deliver information on COVID-19 and Opioid Avoidance. We plan to expand this experience with Watson to other subjects to make the science simple.”

With Watson Assistant and Watson Discovery, you can train your own chatbot at no charge. All you need is a set of documents containing pairs of questions and answers. Watch this how-to video to get started.

The IBM Data Science and AI Elite helps organizations like AdMed adopt AI and tackle top use cases across various industry to help companies — large and small — accelerate their journey to AI. They’re ready to help you kickstart your next data science project.

About the author

Óscar D. Lara-Yejas is Senior Data Scientist with the IBM Data Science and AI Elite and one of the founding members of the IBM Machine Learning Hub. He helps the worlds’ top industries put AI to work, whether it is in healthcare, finance, manufacturing, government, or retail, just to mention a few. He has also contributed to the IBM Big Data portfolio, particularly in the Large-scale Machine Learning area, being an Apache Spark and Apache SystemML contributor.

Óscar holds a Ph.D. in Computer Science and Engineering from University of South Florida. He is the author of the book “Human Activity Recognition: Using Wearable Sensors and Smartphones”, as well as a number of research/technical papers on Big Data, Machine Learning, Human-centric sensing, and Combinatorial Optimization. Dr. Lara-Yejas is a regular speaker at Hadoop Summit, IEEE Congress on Big Data, IEEE World Conference on Computational Intelligence, Strata+Hadoop, IBM Insight, IBM Think, among other conferences and events.