The Austrians have developed a voice-controlled assistant for the maintenance of plant and machinery – based on Professor Hochreiter’s theoretical research into the LSMT algorithm, which was developed in Munich in the 1990s and now forms the basis for Alexa, Cortana and Siri.

“The maintenance engineer talks to the machine,” explains Schwärzler, who studied production management in Vienna. The intelligence of the Workheld solution resides in an inconspicuous tablet. The technician approaches the machine, the machine recognizes the tablet and the conversation starts. “It could sound like this: ‘There are problems with the spindle in the Y-axis of machine no. 5’. The system then searches for the malfunctions and comes up with an answer such as: ‘This same problem occurred two years ago’. It suggests solutions and also tells us who rectified the malfunction at that time. So you can immediately turn to the right colleague who is already familiar with the problem.”

The engineer dictates the test report

In a different scenario the machine reports problems with a pump and the software immediately provides the technician with assembly diagrams and/or searches the database for the experiences gained by other colleagues. The repair orders are fed into an IoT system, which forms the basis for the talking machine. “We are not only problem solvers, but also engage in interactive knowledge management,” says Schwärzler, who founded his company four years ago. The system also records the communication with the technician. “The system remembers customer and project names, assigns information and is constantly expanding its linguistic skills,” Schwärzler adds.

The idea for the “talking machine” came to him and his team via their first product: a classic maintenance tablet containing construction plans and a knowledge database. “We then closely observed our users and quickly noticed that the technicians on site were reluctant to write test reports or documentation,” says Schwärzler. Likewise, out-of-pocket expenses were rarely entered correctly. “There had to be an easier way.” ‘Speech to text’ was the solution – and at the same time a tricky assignment. Today, users can dictate their test reports to the system and report special information directly by voice. Each spoken documentation enriches the solution in terms of content, and other employees or new colleagues benefit from this.

Technology already in use in the car industry

Workheld’s technology is based on well-known voice-controlled assistants such as Alexa and Siri. But the biggest challenge lies in the development of a framework for ‘intent recognition’. In other words, the machine, app, tablet or bot has to understand exactly what the user (e. g., the maintenance technician) wants to achieve. To this end it has to recognize the user’s language input, convert it into text and, if necessary, react to it. “We develop the frameworks for the machines together with our customers on site and deploy various NLP technologies,” Schwärzler explains. NLP stands for Natural Language Processing and describes technologies that are based on machine learning and enable the development of features for understanding natural language in apps, bots and IoT devices. Speech comprehension is the key. Workheld costs €39 per month per user and is available with SAP integration, if required. A large German car manufacturer is already using the technology in collaboration with Workheld.

The main competitors are augmented reality providers. But Workheld can hold its own. The company founder confidently sums up the benefits as follows: “We don’t need a helmet, glasses, large batteries, our eyes don’t get tired, and our hands remain free.” But what about the noise in the factory? Can the technician and the machine understand each other? “We also work with headsets in harsh environments. We have gained good experiences with this,” Schwärzler confirms. And, classically with a ballpoint pen, he enters a task in his to-do list: “I still have to write to Sepp Hochreiter and tell him that we are already working on the talking machine.”