$49.5 billion

value of lab supplies market by 2027

3,461

lab supply manufacturers

7%

anticipated industry growth by 2027

clients_ThermoFisherLogo

EXECUTIVE SUMMARY

TensorIoT helped a research equipment manufacturer create a hands-free voice-assistant tool for use in sterile laboratories.

GOAL

How do we teach Alexa to understand the technical language and complex questions that scientists need answered?

RESULTS

The Tensor team created a hands-free Alexa voice interface that the client’s lab scientists could use to get answers to complex scientific questions.

Background

A leading developer of research equipment wanted to create a connected lab. Many of their products must be produced in a sterile environment, which can be a hassle to maintain. Employees must wear special equipment, regularly disinfect all surfaces, and keep the area free from clutter. With these requirements, tasks as simple as jotting a note down or entering something into a computer can be a hassle. The client wanted to explore whether a hands-free voice-assistant tool could streamline their employee’s workflows.

The Challenge

Building an Alexa tool for scientists was not without its challenges. To start, our Managed AI team needed to familiarize themselves with the specialized chemical terminology and phrases used by the scientists, and then train Alexa to understand them as well. After Alexa had a solid understanding of the scientists’ vocabulary, the Tensor team also needed to teach her to understand the complex questions the scientists needed her to answer, such as the results of chemical conversions. Finally, the team had to train Alexa to recognize and understand a group of users with a diverse range of accents and cultural backgrounds.

The Solution

Using the Amazon Developer Console, the Tensor team created a collection of Alexa skills that taught Alexa how to respond to different questions. To fine-tune the Alexa skills, the team used locale settings to account for different accents. The team also did extensive user testing to capture a variety of phrasing stemming from different cultural backgrounds, to make sure Alexa could understand multiple ways of accomplishing the same intent.