Summary
Before having the help of AI, reading and interpreting a document required using the time of a person, or writing a specialized machine learning model. Semantic Kernel allows you to write code to analyze large and complex documents.
In our pipeline, the CheckSpreadsheet native plugin does not strictly require Semantic Kernel and could be done in a separate step, since it only runs code that is never read by AI. We added it to the pipeline to make our end-to-end solution more streamlined.
The ParseWordDocument native plugin, on the other hand, helps Semantic Kernel receive the information in parts. Breaking the document into parts makes the semantic functions simpler: each function can evaluate just a portion of the document. For example, the function that evaluates the Teams section of the document just needs to check the team qualifications. That makes the function a lot simpler to write than a function that reads the whole document and decides about all sections of...
