Patient enrollment has remained a major challenge for pharmaceutical companies who are conducting clinical trials. Even patients who have the intend to join a trial due to a critical illness finds it extremely difficult to collect information and processes involved in enrolling into a trial. This blog presents the possibilities of building a conversational agent which will mine clinical trial data from centralized portals like clinicaltrials.gov and combine it with a knowledge graph about specific conditions so that patients could seamlessly interact and enroll into a trial.
Our intent is to combine technologies like Crawling, Natural Language Processing and Machine Learning algorithms to extract information about clinical trials and objectively combine them with available knowledge about the condition. The system will also have a dynamic conversational agent which will populate the website with relevant content and guidance as the conversation between the machine and the patient progresses.
This interactive experience is a voice/chat dialogue with a conversational agent that will be embedded into a website. The agent speaks using semantic speech, generated from an seq-seq dialogue model which will be based off of a knowledge graph,. The system will also contain a dynamically-updated user model which will keep the context of the conversations that happened between the patient and the machine.
We plan to integrate conversational nonverbal behavior (hand gestures, facial displays, gaze, etc.) animated in synchrony with the speech. We intend to use this and attract patients with low health literacy and many who have never touched a computer before. The user will be guided through an interactive set of options about their health conditions which could be mapped to the inclusion exclusion criteria thus making the screening process much simpler and agile.