Automated Annotation of Texts using Tactical Cognitive Computing
Across industries, one of the big time sinks is the task of finding some information out of lengthy texts. While this step is critical to any digital workflow, there hardly has been any other way to do it so far except manually. As a result, multitudes of workflows that could benefit from digital technologies do not end up being digitized.
Tactical cognitive computing software like Coseer can be trained to annotate a given document to extract structured information from it. This information is then coded as XML, JSON or other structured formats. For known repetitive patterns, the output is not limited to specific entities or values, but can include abstract ideas repeated throughout the document.
Tactical cognitive computing is uniquely suited for such applications, for the following reasons:
- Usually the context and the information being extracted is very specific to a client and an application. Hence, it is important to train for each situation separately.
- Such steps feed into digital processes. Based on traditional computer science, such processes have limited tolerance for errors. Indeed, usually errors are compounded through such processes. It is very important for success of such workflows to have high accuracy at the outset.
- While such tasks collectively account for a large portion of manual work at many enterprises, individually each task has limited scope. It is important to justify the RoI.
In this blog we present a case study on automated extraction for such information using Coseer’s InfoSeer product. The client in Healthcare industry was able to reduce time taken per document from a week to just a few hours, which included machine assisted verification, and identified significant cost savings.