Gaining competitive advantage from non-textual information
Image processing and computer vision
Wujek indicated that almost every facet of working with such non-textual data is becoming easier for business users, including the data preparation and preprocessing steps for images. Organizations can readily adjust different aspects of images, including their brightness, pixilation, scope of focus, and more, in a self-service, declarative manner. “There are various open source packages certainly that people have developed to do this, so that users don’t have to understand the mechanisms or techniques themselves,” Wujek explained. “They will just say, ‘Smooth this image,’ or, ‘Normalize it,’ or, ‘Impute missing values.’”
The newfound ease for working with visual forms of non-textual data even extends to the data capture process for obtaining the images that inform enterprise knowledge. Organizations can access, fine-tune, build, and deploy computer vision models for applications of object detection, segmentation, and more in a predominantly self-service manner. Contemporary platforms in this space rely on visual, point-and-click, and dragand- drop constructs to perform these necessities.
These capabilities support use cases such as “helping hospitals and medical institutes process their medical images in an automated fashion to identify what appears to be an anomaly,” Wujek said. In that case, practitioners could then evaluate that part of the image, or order additional diagnostic procedures to verify their findings, to see if the image truly illustrates a medical problem. Even when there is coding involved in the model development process for capturing the aforementioned (and other) forms of non-textual content, modern solutions in this space have made it increasingly declarative—in addition to providing out-of-the-box computer vision models. “There are different levels there,” Wujek stipulated. “Some of it is, ‘Hey, build it all yourself.’ Some of it is, ‘Use some of these packaged modules or nodes to piece it together and build it all.’ And some of it is, ‘Hey, we’ve got it all kind of bundled up for you as this model that you can take and apply to your business case directly.’”
The onset of a new age
With self-service constructs to generate descriptions of non-textual information, transform it, and implement it in consuming applications, the value derived from these forms of data will surely increase for KM practitioners. Automatic metadata descriptions, coupled with the tagging and natural language search capabilities of information management systems, will make this content as discoverable, and as useful, perhaps, as textual information is for this discipline. Nonetheless, it is incumbent upon individual organizations to determine how to implement this knowledge so that it delivers competitive advantage—particularly if these capabilities are available to everyone.