Appointments by Conversational AI: Understanding Everything Your Patients Say

78 out of 100 doctors recognize the value of conversational AI for appointment scheduling — Journal of Medical Internet Research

Now, more than ever, patients are looking for ways to communicate with their healthcare providers digitally. With the spread of COVID-19, this demand is only going to increase as health systems across the nation face a tremendous strain on their digital channels and call centers.

Natural Language Processing (NLP), coupled with conversational AI, is poised to become one of the foremost solutions to confront this issue.

The question is, how does it work? 🤔

Last week — as part of #VirtualHIMSS20 and ahead of launching our complimentary COVID-19 Virtual Assistant — we hosted a deep-dive webinar into Hyro’s conversational AI.

We discussed the various benefits and use cases of conversational AI in healthcare and the three pillars behind the Natural Language Understanding that we employ in our solution: Compositionality, Knowledge, and Ambiguity.

Without further ado, let’s recap and revisit some of the webinar’s key insights:

Conversational AI as a Solution

  • Recent studies show that both patients and the health organizations serving them are looking for conversational AI as a means of solving confusion and mitigating patient friction across all digital channels.
  • Most enterprises fail to deploy and maintain conversational interfaces due to high cost, limited scalability, and a lack of NLP/ML expertise.
  • Hyro creates plug & play conversational AI assistants (chat solutions and voice assistants) by seamlessly ingesting information from various data sources.
  • As opposed to typical chat platforms which are based on limited intent-based flows, Hyro uses knowledge graphs to create open “intent-less” conversations.

The Tech Behind Crafting Conversations

Compositionality is a linguistic term that means “putting together.”

  • Intent slots — the predominantly used parameter for compressing intents — fall apart since they fail to accommodate natural language. Sentences are not flat!

Knowledge enables us to model sentences in a compositional way.

  • Data-tables — the standard way of thinking about data representation — fall short when it comes to a. flexibility and b. searching and identifying phrases on vastly-populated data-tables.
  • Classic conversational solutions model conversations as “decision trees” or “state-machines.” User utterances enact the transitions between these “states.”
  • Hyro models conversations as knowledge graphs (networks) rather than tables, which allows our solution to be “schema-less” and flexible.
  • Hyro “enriches” knowledge graphs using websites, databases, and public resources.

Ambiguity is inherent to natural language, as individual words and sentences have different meanings.

  • When choosing a compositional approach for conversational AI, ambiguity is a significant challenge.
  • There are different kinds of ambiguity. The word “may”, is a great example of “token ambiguity” as it could refer to a month, a name, or even a proposition.
  • Hyro uses machine learning models and the patient queries themselves to handle ambiguities.

Modeling language rather than use cases provides ample opportunities for high-level understanding across all use cases, ultimately resulting in better patient experiences.

Converse, Convert, Collect

At the end of the day, our technologies are fuel for that stellar patient journey; when patients can connect and converse easily, they’ll convert. Natural Language allows us to understand, and our knowledge graph enables us to guide — a North Star ultimately navigating patients to achieve their objectives. In fact, 46% more doctor appointments are booked online when patients interact with our conversational assistant. With our dashboard, enterprises can catch those conversations and conversions in real-time and collect insights as to how patients are searching for healthcare, and what elements should be added, revised or removed.

Watch the full webinar here:

…and follow along on Linkedin and Twitter for upcoming announcements & events surrounding NLP, Voice AI and Conversational Health.

We are offering a free COVID-19 virtual assistant to relieve stress on first-line healthcare operations and thwart the spread of fake news, confusion, and misinformation. If you or anyone you know is involved in healthcare, please shoot us a request or learn more at www.hyro.ai/covid-19. The COVID-19 Assistant is free, and we’ll implement it within 48 hours.

Aaron Bours

Head of Marketing at Hyro