
QUANTRON and ETO Motors Announce Collaboration
27. September 2022
The Overseas Port of Rostock is on Course in the First Half of 2022
28. September 2022With the help of artificial intelligence, the Cologne start-up how.fm is elevating its digital training and educational software for warehouse logistics to a new level. A highlight of the new features is the automated extraction and digitization of work instructions from documents.
(Köln) “In the medium term, we want users to be able to independently, flexibly, and without additional expertise take over the structural design for digital training,” explains Farhoud Cheraghi, co-founder and managing director of how.fm. The algorithm for automatic extraction, which the company has trained and developed itself, is an essential component in the development and expansion of how.fm’s digital training software for warehouse workers.
About 15 Percent Time Savings
A product innovation is the automated extraction of existing documentation into training sequences. Using machine learning, relevant work instructions and guidelines are extracted from a PDF file and directly converted into digital sequences for the desired training course. The individual instructions can then be edited as desired or moved to other locations within the course sequence.
Cheraghi clarifies the utility: “The human eye finds it difficult to identify the relevant information in lengthy documents. Often, between 50 to 100 individual instructions are hidden in a PDF. Internal tests have shown that the time savings are about 15 percent compared to the traditional method.” Over the winter, the new feature is set to be rolled out initially with pilot customers and will be available on the market from early 2023.
For Cheraghi, the new feature exemplifies the direction in which the software is to develop: “With our modular principle, users should be able to create and update their individual training programs for onboarding, safety training, or further education without extensive prior knowledge.” Further corresponding features are in planning, such as the automatic derivation of quiz questions from the course sequences or the automated generation of illustration suggestions for individual work instructions.
Refined Recorder App
Another innovation is the optimization of the recorder app, which allows users to create videos themselves. These can be processed more easily in the future by being uploaded directly to the media gallery of the CMS (Content Management System) and integrated into training sequences from there.
Keeping Translations Up-to-Date with AI
A particular challenge is keeping training materials up-to-date in various languages. Especially in the logistics sector, where many employees have a migration background, language barriers are a central problem, explains Cheraghi. His solution: “We have programmed an update path for translations using a multilingual language model that goes far beyond the initial translation. The model is ‘State of the Art’ artificial intelligence in the field of multilingualism.” When changes occur in the production language, the how.fm software automatically recommends whether an update of the translations is necessary. If it involves content or structural corrections, it is suggested to adjust the translations accordingly. However, if a spelling mistake is corrected, the application assesses this as not a significant change. No action is required. The high time and cost expenditure for translations will be significantly reduced through automated selection.
The demand for digital training opportunities is steadily increasing from the managing director’s perspective. A good indicator of the relevance of the digital training software for how.fm is the “healthy usage” of the applications. Since May 2019, the start-up has been measuring how often users access individual work steps. “This October, we will exceed the threshold of one million individual steps executed,” Cheraghi is pleased to announce.
Photo: © how.fm / Image Caption: Managing Director Farhoud Cheraghi






