Voice technology in health research - lots of opportunities and couple of challenges
Voice is not a new method for communicating with computers, and voice controlled machines has been around for a while. IBM shoebox one of the first machines that was capable of understanding voice inputs (see the video below). But today, voice technology have entered a new era of interaction. Through a natural conversational interface, we can maintain human-like conversations with computers.
For the last 5 years, we have been witnessing significant improvements of voice assistants through Google Assistant and Amazon Alexa's conversational skills. With the advancement in AI, voice assistants' capability of maintaining conversation is getting better every day with reduced cost of ownership and an increased adoption rate. It is also scaled-up to be used by billions with Google's deployment of Google Assistant to Android OS mobile phones. On the other hand, the compatibility of voice services with multiple platforms (e.g. IoT and smart home ecosystems) and compliance with regulations (e.g. Amazon Alexa and HIPAA) accelerated the utilization of voiceFirst applications in many industries.
One of the emerging fields utilizing voice has been the healthcare. HIPAA compliant skills and AI powered voiceFirst solutions have been released for enhancing communication in healthcare. Similarly, our team has engaged in couple of funded projects on maternity care and home-based care communication, which helped us to test and implement voiceFirst applications with different healthcare settings. So far, our research showed that, not surprisingly, voice technology is in its early stages (information stage) in health communications. But more importantly, our experiments resulted that the capability of mainstream voice technologies could be limiting evidence-based research.
As the researchers investigating the use of technology in healthcare, we would like to know as much as possible about human-computer interaction (human-voice assistant interaction in this case) to better observe the impact, causality and association with human and computer behavior. However, mainstream voice technologies as Google Assistant and Amazon Alexa's platforms provide access to the data which is limited to pre-defined structured information (like multiple choice responses) and user logs. Practically, the data collected from these platforms are the same as any data that could be collected from a web-based survey. Thus, the core value of voice is still remaining unextracted, which is the conversational data. Not accessing any conversation transcription and audio data limit the voice research in terms of understanding the nature of interaction and decision process during each conversation. Such data would also encapsulate valuable information like voice tone, sentence formation, vocal fluctuations and environmental noise, which could be the indicators of health conditions.
Due to these limitations of mainstream voice technologies, open source alternatives are more appealing for research. Hardware platform like Raspberry Pi is a cheap and versatile alternative for developing custom voice technologies. Mycroft has been in the market which uses Raspberry Pi and open-source developer platform to create voice apps. Jasper is another example for an open source platform that offers development of custom voice commands on custom hardware. These tools could help to create voice interactive ecosystems for home or any other environments that could leverage research in voiceFirst healthcare. Even though, these alternatives are not widely adopted for consumer use, the current efforts with open source voice in healthcare looks promising for research. A good example for leveraging open source voice in public health is the implementation of SickWeather with Mycroft to create a tool for detecting cough in public places to track illness.
At the bottom line, these all are exciting developments, and there are lots of opportunities for research. I will keep on experimenting and exploring mainstream and alternative voice interaction tools and methods in creating voiceFirst healthcare solutions with our team. Stay tuned for more news and researches!
