Business Use case for nlp in vlsi
In any industry there is a QA team. Depending on the type of industry QA teams need to develop test-cases that can be used to test for any defects, standards, profiling of the software/hardware. In VLSI industry this is true as well. Consider a simple scenario of physical verification of chips like DRC. Where we develop a codes to check rules between two geometries. The rule can be as simple as “space between two metal geometries has to be minimum 1um”.
Currently we have to come up with situation where the code might flag false violation or doesn’t flag real violation. Since the rules are written in a specific technical terms it might be possible to automate the task of creating test-cases for each rule by reading (NLP) from a file that contains the description of these rules. Note that there are usually hundreds of rules each requiring a set of test-cases. And these rules vary for different nodes. Achieving this would greatly reduced QA time and development costs.