Towards Extraction of Validation Rules from OPC UA Companion Specifications

Abstract

Interoperability is the key to industrial automation. OPC UA aims to provide interoperability between industrial machines at the network, protocol, and semantic layers. In order to achieve it, the implementation of a machine and its nodeset file should comply with the OPC UA standard. However, there are a few areas for improvement in technology to automatically check compliance since the compliance rules/validation rules are expressed in textual format in specification documents. In other words, they are not machine-interpretable. Converting the text-based specification into machine-readable form is a complex problem since each companion specification is domain-specific and written by a set of industry experts from all over the world from diverse language backgrounds. The specifications are majorly unique, with little commonality between them. Because of these reasons, it is challenging to develop a generic information extraction approach to retrieve rules that work for all specifications. In this study, we aim to handle this challenge to extract the textual rules from specifications automatically by applying Natural Language Processing and Machine Learning technologies. We present our methodology based on Named Entity Recognition for OPC UA documents for information extraction and text classification to identify rules. To achieve this goal, we created five named entity data sets from five selected OPC UA companion specifications. The sentences in the data set and the entities in the sentences were annotated by two OPC UA experts. We point out a deeper analysis of the data sets to highlight common and unique entities in them and show their usage in identifying the rules in the specifications.